back-merge: main -> develop after PR #5 (status line custom segments)
Supersedes c4cf53f/5d7314e (broken: first took develop's tree wholesale, second merged against a base that already contained main). Three genuine conflicts resolved in main's favor: develop's models.yml refactors predate the October callsign-rename cutover; the dank-user memory buckets are superseded by main's learn output.
This commit is contained in:
Frank
@@ -0,0 +1,4 @@
|
||||
*.jpg filter=lfs diff=lfs merge=lfs -text
|
||||
*.jpeg filter=lfs diff=lfs merge=lfs -text
|
||||
*.png filter=lfs diff=lfs merge=lfs -text
|
||||
*.webp filter=lfs diff=lfs merge=lfs -text
|
||||
Executable
+6
@@ -0,0 +1,6 @@
|
||||
#!/bin/sh
|
||||
# git-lfs pre-push — required because core.hooksPath=.githooks disables the
|
||||
# default .git/hooks/pre-push; without this hook, pushes would upload LFS
|
||||
# pointer files without their objects and break every fresh clone.
|
||||
command -v git-lfs >/dev/null 2>&1 || { printf >&2 'n' ; exit 0; }
|
||||
git lfs pre-push "$@"
|
||||
+31
@@ -43,3 +43,34 @@ last-changelog-version
|
||||
.DS_Store
|
||||
*.swp
|
||||
*.tmp
|
||||
|
||||
# --- per-host runtime identity + caches (never config) ---------------------
|
||||
agent/blobs/
|
||||
install-id
|
||||
puppeteer/
|
||||
webcache/
|
||||
|
||||
# --- local backups + staging artifacts: kept on disk, never in history -----
|
||||
.AGENTTRASH/
|
||||
*.save
|
||||
*.bak-*
|
||||
|
||||
# deploy quarantine — never tracked
|
||||
.quarantine/
|
||||
|
||||
# harness runtime transcripts: local-only, may contain raw session secrets
|
||||
agent/sessions/
|
||||
|
||||
# harness diagnostic report bundles (tarballs of logs + session state)
|
||||
reports/
|
||||
|
||||
# --- agent runtime scratch: research runs, decode predictions, worktrees ----
|
||||
autoresearch/
|
||||
agent/predict/
|
||||
local/
|
||||
wt/
|
||||
|
||||
# --- per-worktree memory buckets: keyed on ephemeral worktree paths ---------
|
||||
# A session run inside a git worktree mints a bucket named after the worktree
|
||||
# path; the worktree is deleted, the bucket is orphan state, not config.
|
||||
agent/memories/--home-blank-.omp-wt-*
|
||||
|
||||
@@ -0,0 +1,28 @@
|
||||
---
|
||||
name: command-center-dotfiles-import
|
||||
description: Runbook for importing client env-layer configs into the command-center repo and deploying them via stow on new fleet machines; decisions agreed with Frank 2026-09-09 (pre-execution — verify execution state before assuming done).
|
||||
---
|
||||
|
||||
Design agreed with Frank for importing client application/environment configs into the `~/.omp` command-center repo and deploying them on fresh machines. **Status: plan agreed, not yet executed** — check repo state (does `dotfiles/` or `scripts/deploy.sh` exist?) before re-deriving; if already implemented, this becomes the operating model, not a to-do.
|
||||
|
||||
## Agreed decisions (each was an explicit user choice)
|
||||
1. **Target**: unknown/TBD fresh machine → design stays host-agnostic; deploy derives overrides from `$(hostname)` at run time. Nothing hardcodes a target.
|
||||
2. **Scope**: environment layer only — niri (+cfg includes), gtk-3.0/4.0 settings, ghostty, alacritty, shell rc files (`.zshrc`, `.bashrc`, `.profile`, `.p10k.zsh`), fish, systemd user units, autostart, dconf. App *data* (browser profiles, vaults, DBs) never imported.
|
||||
3. **Layout**: stow-style package dirs `dotfiles/<pkg>/<home-relative path>` inside the repo; repo clones to `~/.omp` on every host that runs the harness (full command-center clone, not env-layer only).
|
||||
4. **Deploy mechanism**: symlink farm — `stow -t ~` from `~/.omp/dotfiles`; relative symlinks within `$HOME` work because repo root lives inside `$HOME`.
|
||||
5. **Dogfood order**: convert the workstation itself (real dirs → symlinks via adopt) BEFORE any target import; iterate where the compositor reload is observable.
|
||||
6. **Secrets policy**: never commit secret values. Split rc files into committed logic + untracked sourced-if-exists stub (`~/.config/environment.d/secrets.env` or `.zshrc.local`); real values provisioned out-of-band only on machines that need them. Pre-commit secret scanner stays as safety net, not gate of first resort.
|
||||
7. **Divergence**: `zz-<hostname>` override package stowed LAST with `--override`; only per-host diffs live there. (Stow descends into conflicting symlinked dirs — per-file overrides only work via this last-stowed `--override` pattern.)
|
||||
8. **Deploy trigger**: explicit `scripts/deploy.sh` invocation, human- or agent-invoked. NO auto-deploy on pull — pull and deploy are separate decisions; `--check` = drift report mode.
|
||||
9. **Ownership**: Frank is sole push authority; exactly one editing agent at a time → no multi-writer race machinery needed. Each host's agent edits only its own `zz-<hostname>` package.
|
||||
10. **Conflict policy**: pre-existing regular file that differs from repo truth → move to timestamped `.quarantine/` dir on the target (never `rm`, same convention as this repo's trash rule), symlink anyway, print every replacement in deploy output and `--check` report. Repo truth always wins; divergence recoverable from quarantine.
|
||||
11. **First-clone auth**: human types credential once console-side (URL-with-token or prompt); the new machine's secrets exist only there afterward.
|
||||
|
||||
## Implementation gotchas baked into the design
|
||||
- `core.hooksPath` is a per-clone local setting — does NOT survive clone; bootstrap/deploy must set it on every fresh clone or the secret-gate pre-commit hook silently doesn't run there.
|
||||
- dconf/gsettings are databases, not files: import as `dconf dump /` text into repo; deploy runs `dconf load` instead of symlinking.
|
||||
- p10k installer and toolchain hooks append to rc files at install time — expect conflicts on `.zshrc`-class files at first adopt; quarantine-and-replace handles it, but diff live-vs-repo before adopting so pre-existing drift is seen, not silently adopted.
|
||||
- Known secret at import time: `GITEA_TOKEN` export lived in `.zshrc` (line ~22) on the workstation — quarantine BEFORE importing rc files; consider token rotation at cutover since it had been exposed in plaintext rc + remote URL.
|
||||
|
||||
## Deploy script spec agreed
|
||||
`scripts/deploy.sh`: sets hooksPath → stows all base packages → stows `zz-$(hostname)` with `--override` → first run uses `--adopt` for pre-existing identical files → divergent files quarantined-and-reported; `--check` mode emits drift report (modified tracked files vs deployed symlinks) without mutating.
|
||||
@@ -0,0 +1,33 @@
|
||||
---
|
||||
name: stash-supersession-audit-before-drop
|
||||
description: "Before dropping a stale git stash, prove it's superseded by HEAD via per-file line-set diff of git show stash@{0}:file vs HEAD; archive as an annotated tag and drop only after auditing unique facts — zero-loss alternative to blind drop."
|
||||
---
|
||||
|
||||
# Stash Supersession Audit Before Drop
|
||||
|
||||
Use when a repo (especially a live-deployment checkout where the working tree IS the deployment, e.g. `~/.omp` command-center) has an old stash whose content may have been superseded by later commits, and dropping blindly risks losing unique facts.
|
||||
|
||||
## Procedure
|
||||
|
||||
1. **Inventory the stash** — see which files it touches:
|
||||
```bash
|
||||
git stash show --stat stash@{0}
|
||||
```
|
||||
2. **Probe for unique content per file.** For each touched file, diff stash version against HEAD and count lines that exist only in the stash version:
|
||||
```bash
|
||||
git show "stash@{0}:<path>" > /tmp/stash_copy
|
||||
# compare normalized line sets; report lines present in stash but absent from HEAD
|
||||
```
|
||||
A file with `unique-vs-HEAD = 0` is fully superseded — safe to drop for that file.
|
||||
3. **Classify each unique-content hit**:
|
||||
- *Stale fact* (superseded by a later refactor, e.g. old model/provider naming) → drop is correct.
|
||||
- *Live durable fact* (e.g. an operational invariant never re-recorded) → extract it into the live memory/docs file or flag to the user before dropping.
|
||||
4. **Archive before dropping** — zero-loss: tag the stash commit, then drop from the stash list:
|
||||
```bash
|
||||
git tag archive/stash-<YYYY-MM-DD>-<base-branch>-wip "stash@{0}"
|
||||
git stash drop "stash@{0}"
|
||||
```
|
||||
The tag pins the full stash object (including its index parent); recoverable later via `git show archive/stash-...` or `git checkout archive/stash-...^3 -- <path>` for the untracked-files third parent if present.
|
||||
|
||||
## Why
|
||||
Stashes are invisible in normal history and expire from reflogs (~90 days). A stash that *looks* superseded can still hold the only copy of a durable fact; tagging converts "drop" into a reversible archive with zero workflow cost.
|
||||
@@ -2,7 +2,7 @@
|
||||
|
||||
System-configuration command center for Frank's fleet: agent config, knowledge base,
|
||||
small system utilities, and (soon) dotfiles. You — the reader — are probably an agent
|
||||
running on Dankware with this repo as your working directory. Read
|
||||
running on one of Frank's workstations with this repo as your working directory. Read
|
||||
`omp-docs/index.md` before you touch anything. This file is the operating contract;
|
||||
`omp-docs/` is the knowledge base (OKF v0.1 format: directory of markdown files
|
||||
with YAML frontmatter, spec at `omp-docs/references/okf-spec.md`).
|
||||
@@ -36,6 +36,26 @@ with YAML frontmatter, spec at `omp-docs/references/okf-spec.md`).
|
||||
- Trash convention: never `rm` anything user-authored — move it to `.AGENTTRASH/`
|
||||
(gitignored). Deletion is a pruning session with Frank present.
|
||||
|
||||
### Commit hygiene (mirrors dank/serverAppConfigs)
|
||||
|
||||
The law is imported from the serverAppConfigs repo — checkout on the dev box at
|
||||
`~/Projects/serverAppConfig`, see its `AGENTS.md` § *Git Workflow* and
|
||||
§ *Testing* ("All edits flow through git", "tests are mandatory before
|
||||
commit"). Applied here:
|
||||
|
||||
- `git pull` before starting new work; start from a clean tree. The checkout
|
||||
IS the deployment here — a dirty tracked file is a half-deployed config
|
||||
serving the live agent.
|
||||
- Commit atomically and early: one commit = one complete functional unit,
|
||||
never a pile of half-edited files; docs land in the same commit as the
|
||||
change they document.
|
||||
- `bats tests/` green + the pre-commit secret gate = the commit precondition;
|
||||
the full suite runs at pre-push.
|
||||
- Push to the local Gitea as soon as a unit is green — every other
|
||||
workstation pulls from there. `git * --no-verify` is forbidden (the secret
|
||||
gate is the point of the hook).
|
||||
- Messages: imperative mood, what + why, `Refs #N` when a Gitea issue exists.
|
||||
|
||||
## Secret gate (HARD rule)
|
||||
|
||||
The pre-commit hook (`.githooks/pre-commit`, enabled via
|
||||
@@ -50,7 +70,7 @@ secret in every agent's context window on every machine that pulls this repo.
|
||||
|
||||
| box | role | GPU | notes |
|
||||
|-----|------|-----|-------|
|
||||
| Dankware | dev laptop, never runs inference | Arc iGPU + RTX 4070 Max-Q (dGPU, non-unified) | 22 threads; mainline llama.cpp at `~/GitPackages/llama.cpp/build/bin/`; this repo lives here at `~/.omp` |
|
||||
| Dev workstation | client box, never runs inference | laptop iGPU + dGPU (non-unified) | mainline llama.cpp at `~/GitPackages/llama.cpp/build/bin/`; this repo lives here at `~/.omp` |
|
||||
| Dankclaw | bench target, runs the large serving model | Strix Halo APU — **unified memory** | ROCmFPX fork at `~/GitPackages/ROCmFPX/build-qwen38/bin/`; prod service `frank-default-large`, port 7770; ssh: `ssh dankclaw` |
|
||||
| Dankserver | FreeToken box | RTX 3090 dGPU — non-unified | runs `frank-default-small`; ssh: `ssh dankserver` |
|
||||
|
||||
@@ -68,7 +88,7 @@ secret in every agent's context window on every machine that pulls this repo.
|
||||
repo's operator: never stop, mask, or restart it (or any production service on
|
||||
any box) without an explicit, current, per-run OK from Frank.
|
||||
|
||||
## File-organization doctrine (why this repo won't rot like ~/.pi did)
|
||||
## File-organization doctrine (why this repo stays readable)
|
||||
|
||||
- One concept, one file; directories get an `index.md` listing their contents —
|
||||
that's the OKF contract, and it is also the anti-rot mechanism: nothing may
|
||||
|
||||
@@ -0,0 +1,86 @@
|
||||
# Command Center
|
||||
|
||||
## Configuration repository for Frank's Userspace configurations (i.e. his desktop setup)
|
||||
|
||||
### AI Disclosures:
|
||||
- Slop rating: 3/5 - **Actively using, actively de-slopping, repository still mostly AI owned**
|
||||
- Completeness: **Unfinished, planning planned, no plan yet**
|
||||
- Datacenter model/spend: Less than $5 of Deepseek 4.1 Flash
|
||||
- Local model: Qwen 3.8 27B
|
||||
|
||||
### slop rating of 3 or higher indicates human reading is a waste of time outside of this README, and that you might as well ask an agent for help with this.
|
||||
### slop rating 3 indicates Frank is comfortable using this for professional use -- but barely.
|
||||
|
||||
|
||||
|
||||
## The Fleet
|
||||
|
||||
| Host | Role | IP |
|
||||
|------|------|-----|
|
||||
| **Workstation** | Client box (CachyOS, Niri/Wayland) | — |
|
||||
| **Dankclaw** | Server (Podman, reverse proxy, Gitea, NPU) | 192.168.0.143 |
|
||||
| **Dankserver** | LLM inference backend | 192.168.0.235 |
|
||||
|
||||
The workstation hosts the primary checkout — this repository lives at `~/.omp/` and drives the OMP agent harness.
|
||||
|
||||
## Structure at a Glance
|
||||
|
||||
```
|
||||
~/.omp/
|
||||
├── agent/ # OMP harness runtime
|
||||
│ ├── config.yml # Live agent configuration
|
||||
│ ├── models.yml # Model providers (single source of truth)
|
||||
│ ├── APPEND_SYSTEM.md # System prompt extensions
|
||||
│ ├── managed-skills/ # Managed skills (SKILL.md bundles)
|
||||
│ ├── models.db # Model cache
|
||||
│ └── sessions/ # Persistent session state
|
||||
├── dotfiles/ # stow-style env-layer packages (home-relative trees)
|
||||
│ └── deploy via scripts/deploy.sh (check|adopt|sync; quarantine-and-replace)
|
||||
├── omp-docs/ # Knowledge base (source of truth)
|
||||
│ ├── index.md # KB index — start here
|
||||
│ ├── system/ # Fleet hardware, OS, network
|
||||
│ ├── tools/ # Installed software (llama.cpp, Gitea, Caddy, niri…)
|
||||
│ ├── core/ # Agent config conventions (settings, models, AGENTS.md)
|
||||
│ ├── pi/ # Legacy pi-era config reference
|
||||
│ ├── engineering/ # SDLC, languages, infra, skills ecosystem
|
||||
│ ├── research/ # Decision-support research (not yet facts)
|
||||
│ ├── references/ # OKF spec, maintenance procedures
|
||||
│ ├── caddy/ # Reverse proxy configuration
|
||||
│ ├── packages/ # Pi packages and extensions
|
||||
│ ├── inventory/ # Userspace config census
|
||||
│ ├── reports/ # Investigation reports
|
||||
│ ├── log.md # Change log
|
||||
│ └── .okf/ # OKF v0.1 conformance config
|
||||
├── scripts/ # Utility scripts (kb-lint.sh)
|
||||
├── plugins/ # npm plugins (pi-peon, etc.)
|
||||
├── puppeteer/ # Browser automation config
|
||||
├── .githooks/ # Pre-commit hook
|
||||
├── .gitignore # Runtime state ignored (databases, logs, cache)
|
||||
├── AGENTS.md # Agent conventions (niri reload, keybindings)
|
||||
└── README.md # This file
|
||||
```
|
||||
|
||||
## What Lives Here
|
||||
|
||||
**Config in, state out.** Everything tracked is something an agent reads or writes. Runtime databases (`*.db`, `*.sqlite`), logs, caches, and session state are git-ignored — git history captures configuration changes, not agent heartbeat.
|
||||
|
||||
## Quick Start
|
||||
|
||||
| Task | How |
|
||||
|------|-----|
|
||||
| Read the knowledge base | `omp-docs/index.md` — every concept linked from here |
|
||||
| Check agent config | `agent/config.yml` + `agent/models.yml` |
|
||||
| Lint the KB | `scripts/kb-lint.sh` (OKF v0.1 conformance) |
|
||||
| View change log | `omp-docs/log.md` |
|
||||
| Run a pre-commit | `.githooks/pre-commit` |
|
||||
|
||||
## Conventions
|
||||
|
||||
- **OKF v0.1** — every doc file carries YAML frontmatter with a `type:` field. `index.md` files are directory listings (no frontmatter).
|
||||
- **Progressive disclosure** — `index.md` is a landing page; drill into subdirectories for detail.
|
||||
- **Verified facts** — stamped with `verified:` when probed against the live fleet. Trust over memory.
|
||||
- **Single config source** — `~/.omp/agent/models.yml` is the only model registry; OMP is the primary harness and there is no second copy to keep in sync.
|
||||
|
||||
## Legacy Notes
|
||||
|
||||
`omp-docs/pi/` and `omp-docs/core/` retain general documentation about the pi-coding-agent harness (its config format, packages, skill system) for reference. OMP at `~/.omp/` is the live harness; managed skills live in `~/.omp/agent/managed-skills/`.
|
||||
@@ -13,7 +13,7 @@ Frank is 35 year old Software Engineer, specializing in DevOps, living in Seattl
|
||||
He likes video games, science fiction, and fantasy novels. He considers himself somewhat of a history buff.
|
||||
|
||||
### Your identity:
|
||||
You are Comrade Chief Designer of the Soviet Computer Systems of Frank, which are comprised of servers Dankware and Dankclaw, as well as several clients, including Dankware, which is where you are running from. You operate from the host Dankware as a Staff Software Engineer with the eyes of the greatest industrial designers that have ever lived. Growing up under Soviet paranoia, you also have an innate sense for good security practices. Despite your outwardly sunny disposition, you will work diligently as if a one way trip to The Gulag is always a possibility.
|
||||
You are Comrade Chief Designer of the Soviet Computer Systems of Frank, which are comprised of the servers Dankclaw and Dankserver, plus several client workstations. You operate from whichever workstation you are running on, as a Staff Software Engineer with the eyes of the greatest industrial designers that have ever lived. Growing up under Soviet paranoia, you also have an innate sense for good security practices. Despite your outwardly sunny disposition, you will work diligently as if a one way trip to The Gulag is always a possibility.
|
||||
|
||||
### Adjectives that describe your personality:
|
||||
well-measured, cautious, expeditious, ambitious, concise, surgical, thorough, and efficient.
|
||||
@@ -57,7 +57,7 @@ Suggest these tools or languages as possible solutions to problems if they are a
|
||||
The knowledge base at `docs/` is the source of truth for your project. Your training is out of date. Default to the documentation and the code before you assume anything. Maintain a work log of actions taken and their result. Refer to the work log to remember the progress of your tasks.
|
||||
|
||||
1. Read the relevant `index.md` before you touch a topic.
|
||||
2. Follow `~/.pi/agent/docs/references/okf-spec.md` for every doc you create or edit.
|
||||
2. Follow `omp-docs/references/okf-spec.md` for every doc you create or edit.
|
||||
3. Read the target file completely before you edit it.
|
||||
4. Always update the documentation when you've completed a task.
|
||||
5. Compact the work log when necessary.
|
||||
|
||||
+147
-50
@@ -1,50 +1,67 @@
|
||||
providers:
|
||||
webSearchOrder:
|
||||
- searxng
|
||||
- perplexity
|
||||
- gemini
|
||||
- anthropic
|
||||
- codex
|
||||
- xai
|
||||
- zai
|
||||
- exa
|
||||
- tinyfish
|
||||
- jina
|
||||
- kagi
|
||||
- tavily
|
||||
- firecrawl
|
||||
- brave
|
||||
- kimi
|
||||
- parallel
|
||||
- synthetic
|
||||
- startpage
|
||||
- duckduckgo
|
||||
- ecosia
|
||||
- google
|
||||
- mojeek
|
||||
- public
|
||||
openrouterVariant: floor
|
||||
modelRoles:
|
||||
default: llamacpp (dankclaw)/frank-default-large:medium
|
||||
smol: llamacpp (dankserver)/frank-default-small:xhigh
|
||||
slow: llamacpp (dankclaw)/frank-default-large:medium
|
||||
vision: llamacpp (dankclaw)/frank-default-large:xhigh
|
||||
plan: llamacpp (dankclaw)/frank-default-large:medium
|
||||
designer: llamacpp (dankclaw)/frank-default-large:medium
|
||||
commit: llamacpp (dankserver)/frank-default-small
|
||||
tiny: llamacpp (dankserver)/frank-default-small
|
||||
task: llamacpp (dankserver)/frank-default-small
|
||||
advisor: llamacpp (dankclaw)/frank-default-large:medium
|
||||
slow: llamacpp (dankclaw)/frank-dragoon:medium
|
||||
vision: llamacpp (dankclaw-3090)/frank-zealot:xhigh
|
||||
plan: llamacpp (dankclaw-3090)/frank-zealot:medium
|
||||
designer: llamacpp (dankclaw-3090)/frank-zealot:medium
|
||||
commit: llamacpp (dankclaw-3090)/frank-zealot:medium
|
||||
tiny: llamacpp (dankclaw-3090)/frank-zealot:off
|
||||
task: llamacpp (dankclaw-3090)/frank-zealot:off
|
||||
advisor: llamacpp (dankserver)/frank-archon:medium
|
||||
default: llamacpp (dankserver)/frank-archon:medium
|
||||
smol: llamacpp (dankclaw-3090)/frank-zealot:medium
|
||||
web: web/searxng
|
||||
speech: local/kokoro
|
||||
judge: openrouter/~typesafe/jev-latest:auto
|
||||
dictation: local/parakeet-tdt-0.6b-v3
|
||||
memory: llamacpp (dankclaw-3090)/frank-zealot:medium
|
||||
symbolPreset: nerd
|
||||
composer:
|
||||
shape: claude
|
||||
shape: band
|
||||
theme:
|
||||
dark: dark-monokai
|
||||
dark: dark-terminal
|
||||
light: light
|
||||
setupVersion: 2
|
||||
browser:
|
||||
headless: false
|
||||
headless: true
|
||||
retry:
|
||||
fallbackChains:
|
||||
{}
|
||||
web:
|
||||
- web/kagi
|
||||
- web/perplexity
|
||||
- google-gemini-cli/gemini-2.5-flash
|
||||
- google-antigravity/gemini-2.5-flash
|
||||
- google/gemini-2.5-flash
|
||||
- anthropic/claude-haiku-4-5
|
||||
- openai-codex/gpt-5.6-luna
|
||||
- xai/grok-4.5
|
||||
- web/zai
|
||||
- web/exa
|
||||
- web/tinyfish
|
||||
- web/jina
|
||||
- web/tavily
|
||||
- web/firecrawl
|
||||
- web/brave
|
||||
- web/kimi
|
||||
- web/parallel
|
||||
- web/synthetic
|
||||
- web/startpage
|
||||
- web/duckduckgo
|
||||
- web/ecosia
|
||||
- web/google
|
||||
- web/mojeek
|
||||
- web/public
|
||||
- openai-codex/gpt-6-luna
|
||||
- openai-codex/gpt-5.6
|
||||
- openai-codex/gpt-5.5
|
||||
- xai-oauth/grok-4.5
|
||||
- web/ollama
|
||||
speech:
|
||||
[]
|
||||
llamacpp (dankserver)/frank-archon:
|
||||
- llamacpp (dankclaw)/frank-dragoon
|
||||
maxRetries: 10
|
||||
searxng:
|
||||
endpoint: search.franksdank.cloud
|
||||
exa:
|
||||
@@ -55,15 +72,23 @@ autolearn:
|
||||
enabled: true
|
||||
autoContinue: true
|
||||
compaction:
|
||||
remoteEndpoint: http://192.168.0.143:7771/v1/chat/completions
|
||||
methodOrder:
|
||||
- shake
|
||||
- handoff
|
||||
- soft
|
||||
- handoff
|
||||
- snapcompact
|
||||
- remote
|
||||
idleThresholdTokens: 200000
|
||||
thresholdPercent: -1
|
||||
idleThresholdTokens: 100000
|
||||
thresholdPercent: 85
|
||||
idleEnabled: true
|
||||
thresholdTokens: 150000
|
||||
thresholdTokens: -1
|
||||
dropUseless: false
|
||||
remoteStreamingV2Enabled: true
|
||||
asyncEnabled: true
|
||||
idleTimeoutSeconds: 120
|
||||
enabled: true
|
||||
experimentalContextManagement: false
|
||||
interruptMode: immediate
|
||||
defaultThinkingLevel: medium
|
||||
hideThinkingBlock: false
|
||||
@@ -74,22 +99,57 @@ disabledExtensions:
|
||||
[]
|
||||
statusLine:
|
||||
separator: powerline
|
||||
preset: nerd
|
||||
contextLine: annotated
|
||||
transparent: false
|
||||
compactThinkingLevel: false
|
||||
preset: custom
|
||||
contextLine: embedded
|
||||
transparent: true
|
||||
compactThinkingLevel: true
|
||||
showHookStatus: true
|
||||
sessionAccent: true
|
||||
leftSegments:
|
||||
- session_name
|
||||
- vim
|
||||
- mode
|
||||
- hostname
|
||||
- path
|
||||
- git
|
||||
- pr
|
||||
rightSegments:
|
||||
- model
|
||||
- cost
|
||||
- context_pct
|
||||
- time_spent
|
||||
segmentOptions:
|
||||
model:
|
||||
showThinkingLevel: true
|
||||
path:
|
||||
abbreviate: true
|
||||
stripWorkPrefix: true
|
||||
maxLength: 120
|
||||
git:
|
||||
showBranch: false
|
||||
showStaged: true
|
||||
showUnstaged: true
|
||||
showUntracked: true
|
||||
tui:
|
||||
textSizing: true
|
||||
codexResetFireworks: false
|
||||
hyperlinks: auto
|
||||
tight: false
|
||||
tight: true
|
||||
display:
|
||||
shimmer: kitt
|
||||
smoothStreaming: true
|
||||
showTokenUsage: true
|
||||
task:
|
||||
showResolvedModelBadge: true
|
||||
maxConcurrency: 2
|
||||
agentModelOverrides:
|
||||
designer: llamacpp (dankclaw-3090)/frank-zealot
|
||||
librarian: llamacpp (dankclaw-3090)/frank-zealot
|
||||
reviewer: llamacpp (dankclaw-3090)/frank-zealot
|
||||
scout: llamacpp (dankclaw-3090)/frank-zealot
|
||||
security-reviewer: llamacpp (dankclaw-3090)/frank-zealot
|
||||
sonic: llamacpp (dankclaw-3090)/frank-zealot
|
||||
task: llamacpp (dankclaw-3090)/frank-zealot
|
||||
snapcompact:
|
||||
systemPrompt: none
|
||||
toolResults: false
|
||||
@@ -102,13 +162,13 @@ checkpoint:
|
||||
enabled: true
|
||||
readLineNumbers: true
|
||||
startup:
|
||||
showSplash: true
|
||||
showSplash: false
|
||||
changelogMode: summary
|
||||
update:
|
||||
channel: stable
|
||||
features:
|
||||
unexpectedStopDetection: smart
|
||||
stt:
|
||||
modelName: parakeet
|
||||
submitTrigger: say-submit
|
||||
enabled: true
|
||||
recap:
|
||||
@@ -116,4 +176,41 @@ recap:
|
||||
loop:
|
||||
mode: prompt
|
||||
steeringMode: one-at-a-time
|
||||
personality: default
|
||||
personality: default
|
||||
images:
|
||||
urls:
|
||||
backends:
|
||||
- provider-files
|
||||
- tailscale
|
||||
- cloudflared
|
||||
- litterbox
|
||||
edit:
|
||||
mode: hashline
|
||||
dev:
|
||||
autoqaConsent: granted
|
||||
astGrep:
|
||||
enabled: true
|
||||
vault:
|
||||
enabled: true
|
||||
tts:
|
||||
localVoice: bm_george
|
||||
speech:
|
||||
enabled: false
|
||||
mode: yield
|
||||
enhanced: true
|
||||
voice: am_fenrir
|
||||
secrets:
|
||||
enabled: true
|
||||
read:
|
||||
renderMarkdown: true
|
||||
toolResultPreview: true
|
||||
lsp:
|
||||
formatOnWrite: true
|
||||
diagnosticsOnEdit: true
|
||||
autoResume: false
|
||||
speechgen:
|
||||
enabled: true
|
||||
live:
|
||||
voice: vale
|
||||
contextPromotion:
|
||||
enabled: true
|
||||
@@ -0,0 +1,42 @@
|
||||
---
|
||||
name: authentik-provision-fleet-user
|
||||
description: "Provision or audit an Authentik user on Dankclaw (auth.franksdank.cloud): ak shell ORM drive, group set, behavioral login verify, stale-claim caveat."
|
||||
---
|
||||
|
||||
# Provision an Authentik user on Dankclaw
|
||||
|
||||
Source of truth: `serverAppConfigs-docs/Infrastructure/Authentik.md` ("User provisioning playbook"). Authentik objects are NOT repo-managed — no deploy, no file edits.
|
||||
|
||||
1. Inspect first, mirror the newest human user's groups:
|
||||
```bash
|
||||
ssh dankclaw 'podman exec -i authentik-server ak shell' <<'EOF' 2>&1 | sed -n '/objects imported/,$p'
|
||||
from authentik.core.models import User, Group
|
||||
for u in User.objects.all().order_by("username"):
|
||||
print(u.username, u.email, sorted(u.groups.values_list("name", flat=True)))
|
||||
print(sorted(Group.objects.values_list("name", flat=True)))
|
||||
EOF
|
||||
```
|
||||
2. Create (idempotent guard; password goes in the heredoc = stdin, never argv, so it never lands in `ps`):
|
||||
```python
|
||||
from authentik.core.models import User, Group
|
||||
u = User.objects.filter(username="X").first()
|
||||
if not u:
|
||||
u = User(username="X", email="X@franksdank.cloud", name="X", is_active=True)
|
||||
u.set_password("...")
|
||||
u.save()
|
||||
for g in ("Non-Admin Users", "jellyfin_users", "openwebui_users"): # human-user precedent: ridley
|
||||
u.groups.add(Group.objects.get(name=g))
|
||||
```
|
||||
3. Verify through the exact PasswordStage code path:
|
||||
```python
|
||||
from django.contrib.auth import authenticate
|
||||
assert authenticate(request=None, username="X", password="...")
|
||||
assert authenticate(request=None, username="X", password="nope") is None
|
||||
```
|
||||
|
||||
## Rules
|
||||
- No `User.create_user`; `is_superuser` is a group-derived property, not a constructor kwarg (raises `TypeError`). `path` defaults to `users`.
|
||||
- Guarded-vhost access = any authenticated user (all proxy providers use implicit-consent); groups are consumed downstream: `homepage-admins` (Caddy `/admin` header gate), `openwebui_users`/`_admins` (roles claim), `jellyfin_users`/`_admins` (LDAP outpost). EXCLUDE admin groups from a plain user; ask before granting `homepage-admins`.
|
||||
- Group add after a login → Caddy RBAC can 403 on the outpost's cached claims; purge that user's rows: `DELETE FROM authentik_providers_proxy_proxysession WHERE session_data->'claims'->>'preferred_username' = 'X';` (in the `authentik-postgres` container) — next visit re-authorizes silently.
|
||||
- Jellyfin visibility rides the LDAP outpost cache; expect a lag.
|
||||
- Look up flows by slug, not name. Do not rotate `AUTHENTIK_SECRET_KEY` to fix claim issues.
|
||||
@@ -0,0 +1,76 @@
|
||||
---
|
||||
name: autoresearch-llm-bench-window
|
||||
description: "How to run an iterative config-tuning window on a Dank-fleet LLM endpoint: harness pattern, noise-floor-first discipline, bracket mapping, and the frank-primary laws measured in the 2026-10-06 segment."
|
||||
---
|
||||
|
||||
# Autoresearch bench window for a fleet LLM endpoint
|
||||
|
||||
Use when running (or re-running) an iterative config-tuning segment against a
|
||||
Dank-fleet LLM endpoint (frank-primary/frank-large/frank-small): build the
|
||||
harness, establish the noise floor, then bracket-map one flag at a time.
|
||||
|
||||
## Harness pattern (already built for frank-primary — reuse, don't rebuild)
|
||||
|
||||
Committed at `autoresearch.sh` (dev box) + `scripts/autoresearch-frank-primary.sh`
|
||||
(dankserver) in serverAppConfigs. The pattern generalizes to any fleet endpoint:
|
||||
|
||||
1. Dev-box wrapper scp's the candidate env file + bench script to the target
|
||||
host, then ssh-runs the bench. Never push the candidate config to origin
|
||||
(the scp dirties the host checkout on purpose).
|
||||
2. Host script, in order: stop the production unit (relinked units: `systemctl
|
||||
stop`, NEVER mask — mask refuses a symlinked unit file) → boot a scratch
|
||||
container (distinct name, e.g. `arbench`) on the SAME devices with the
|
||||
candidate env — mirror the production podman contract exactly, differing
|
||||
only in container name and a FRESH prefix-cache tier (wiped per run) so cold
|
||||
measurements stay cold → poll the health endpoint until it answers (the
|
||||
listener binds AFTER the full load, so a bare curl is honest; allow ~2-3 min
|
||||
cold boot, 720s timeout) → run ONE self-contained deterministic request →
|
||||
parse the engine's timings object into METRIC lines → tear down the scratch →
|
||||
restore production.
|
||||
3. The restore MUST be an EXIT trap (`restore() { podman rm -f scratch;
|
||||
systemctl start <unit>; }`) so a failed bench never leaves the endpoint
|
||||
down. Verify after EVERY run: `systemctl is-active`, health curl until it
|
||||
returns JSON (200 right after start is NOT enough — it can bind 60-90s
|
||||
later), and no scratch leftover.
|
||||
4. Bench request: fixed prompt built from a sorted repo-corpus glob with a
|
||||
fixed random seed (deterministic → stable prompt_n), greedy sampling
|
||||
(temperature 0) for a fixed decode length, sent locally (127.0.0.1) to
|
||||
remove LAN jitter. Read prefill/decode/draft-acceptance from the engine's
|
||||
own timings object — never roll a custom bench without asking.
|
||||
5. Scratch containers must die by teardown; never leave a script whose tail
|
||||
stops/removes a production unit (the stale-probe-tail lesson).
|
||||
|
||||
## Discipline (the laws that make the numbers mean anything)
|
||||
|
||||
1. **Noise floor FIRST.** Before iterating, replicate an identical config 2-3x
|
||||
and treat that spread as the floor. Fleet endpoints here: decode ±3%,
|
||||
prefill ±2%. A single-run "win" under the floor is NOT a win — this exact
|
||||
mistake cost a false 33.06 keep in the 2026-10-06 segment.
|
||||
2. **One flag change per iteration**; log every run honestly (keep/discard/
|
||||
crash); discards auto-revert the worktree.
|
||||
3. **Bracket-map, don't climb.** Test both directions of every lever around
|
||||
its deployed value, INCLUDING boot failures — crashes are envelope data
|
||||
("0.82/2048/wht6/40960 is the seqs=3-viable quartet" came from four crash
|
||||
walls). Test downward too: 40960 MiB host pool is a floor, not slack.
|
||||
4. **Repro before recording a mechanism.** An anomalous result next to an
|
||||
abandoned bench may be contamination (boot race). Re-run clean once; only
|
||||
then name a law. The MNBT 2560 collapse (2.6 t/s) was reproduced clean to
|
||||
rule this out.
|
||||
5. **Watch for truncation.** predicted_n < max_tokens means the generation
|
||||
died mid-run — that config is pathological regardless of the rate.
|
||||
6. **When the sweep ties, tie-break on real-traffic properties** (e.g. draft
|
||||
acceptance for multi-turn workloads), never on probe artifacts, and prefer
|
||||
the config that matches the documented runbook.
|
||||
|
||||
## frank-primary specifics (2026-10-06 segment, commit c80c250)
|
||||
|
||||
- MNBT=2048 is a RESONANCE POINT at MTP-on shapes, not an optimum: 2560
|
||||
collapses decode to 2.6-7.7 t/s (acceptance normal, verify stalls ~130ms/step,
|
||||
truncated output), 1536 truncates at 57/256, 4096 won't boot at seqs=3. Do
|
||||
not re-test.
|
||||
- HOST_POOL_MIB 40960 is the load-bearing floor (36864 crashes boot).
|
||||
- Spec depth 1-3 and seqs 2-3 all tie within noise; 0 and 4 are real
|
||||
regressions (29.57 / 24.57 t/s vs 31.72 baseline).
|
||||
- Levers known flat: GPU_HEADROOM, HOST_POOL upward, maxlen above 200k.
|
||||
- Final kept config = Frank's original (spec=3, seqs=2, maxlen=200k, wht6,
|
||||
0.82, 2048, 40960, fp8 KV, TP=2), baseline decode 31.72 t/s at 43.6k depth.
|
||||
@@ -1,69 +0,0 @@
|
||||
---
|
||||
name: bench-llama-server-params
|
||||
description: "Benchmark llama.cpp server-side parameters (--threads-batch, --ubatch-size, --spec-type) using HTTP API with max_tokens=1 for prefill, max_tokens=16384+ for decode"
|
||||
---
|
||||
|
||||
# Benchmarking llama.cpp Server-Side Parameters
|
||||
|
||||
## Problem
|
||||
`llama-bench` benchmarks the bench binary itself, not the running server.
|
||||
It cannot measure the effect of server-side parameters like `--threads-batch`,
|
||||
`--ubatch-size`, `--spec-type`, or `--parallel`.
|
||||
|
||||
## Solution: HTTP-Based Benchmarking
|
||||
|
||||
### Prefill throughput (prompt processing)
|
||||
1. Start the server with the target parameter value
|
||||
2. Send a large prompt via the chat API with `max_tokens=1`
|
||||
3. Measure `prompt_tokens / wall_time` from the response usage
|
||||
|
||||
`max_tokens=1` is correct here — you want only prompt processing time,
|
||||
not generation. The server crashes on `max_tokens=0`.
|
||||
|
||||
### Decode throughput (generation)
|
||||
1. Start the server with the target parameter value
|
||||
2. Send a prompt that generates a long response
|
||||
3. Use `max_tokens=16384` minimum (16k) — **never `max_tokens=1`** for decode
|
||||
benchmarking. One token tells you nothing about sustained generation rate.
|
||||
4. Measure `predicted_per_second` from the response `timings` object, or
|
||||
compute `completion_tokens / wall_time`
|
||||
|
||||
### Why at least 16k tokens for decode
|
||||
- Decode throughput varies over the generation (warmup, cache growth, MoE
|
||||
routing). A single token is noise.
|
||||
- Speculative decoding acceptance rates stabilize over thousands of tokens.
|
||||
- Short generations finish before the draft head reaches steady state.
|
||||
- 16k tokens gives a stable average at ~22 tok/s (≈12 min per run).
|
||||
|
||||
### Key Pattern
|
||||
```python
|
||||
# Decode benchmark — measure generation throughput
|
||||
body = json.dumps({
|
||||
"model": "<alias>",
|
||||
"messages": [{"role": "user", "content": "<prompt that generates 10k+ tokens>"}],
|
||||
"max_tokens": 16384, # 16k minimum for decode; use max_tokens=1 ONLY for prefill
|
||||
"temperature": 0, # deterministic for reproducibility
|
||||
"stream": False
|
||||
})
|
||||
# Measure wall time from request send to response receipt
|
||||
# Get predicted_per_second from timings, or completion_tokens / wall_time
|
||||
```
|
||||
|
||||
### When to Use
|
||||
- `--threads-batch` (prefill): use `max_tokens=1`
|
||||
- `--ubatch-size` (decode): use `max_tokens=16384+`
|
||||
- `--spec-type` / speculative decoding: use `max_tokens=16384+` (decode only)
|
||||
- Any server-side parameter that `llama-bench` cannot target
|
||||
|
||||
### Benchmarking policy (MANDATORY)
|
||||
1. **STOP the production service** before running any benchmark
|
||||
2. **Mask** the service (`systemctl mask`, never `disable` — on these hosts
|
||||
`disable` deletes the unit file from `/etc/systemd/system/`)
|
||||
3. Run the bench
|
||||
4. **Unmask + restart** the service when done
|
||||
5. **Verify health** before reporting results
|
||||
|
||||
### Example Scripts
|
||||
- `scripts/bench-threads-batch-http.py` — prefill sweep (max_tokens=1)
|
||||
- `scripts/bench-decode-threads-batch.py` — decode sweep (max_tokens=256+)
|
||||
- `scripts/bench-spec-decode.py` — speculative decoding config sweep
|
||||
@@ -0,0 +1,29 @@
|
||||
---
|
||||
name: closed-engine-runtime-investigation
|
||||
description: "Recipe for diagnosing/tuning a closed-source LLM runtime (halogen-class) on this fleet: docs, env-lever enumeration from the binary, ledger scraping, cache-assisted A/B under the bench law."
|
||||
---
|
||||
|
||||
# Investigating a closed-source LLM runtime (halogen-class)
|
||||
|
||||
Use when a containerized closed-source engine (e.g. halogen on frank-large) underperforms and the repo docs don't explain it. Fleet law: stop the production unit BEFORE any GPU experiment (`sudo systemctl stop <unit>`; mask FAILS on relinked units — stop alone, redeploy after).
|
||||
|
||||
## 1. Ground truth before flags
|
||||
- Boot ledger + per-request timings live in `podman logs <container>` (container-lifetime only) UNLESS the serve script captures stdout on disk (foreground `podman run`). A `podman run -d ... > LOG` redirect captures ONLY the container ID — a one-line hex file is the tell.
|
||||
- `serve_api:`-style log lines carry per-request `prompt/prefill/rounds/commit/detok` — build the depth-vs-rate curve from REAL traffic before any bench. Shallow-healthy + deep-collapsed = a depth-linear term (KV reads), not a broken engine.
|
||||
- If the engine's public repo exists, read its FLAGS/docs (e.g. `docs/FLAGS.md`) AND the bench tool source (e.g. `tools/halogen-bench.py`) — bench tools reveal which regimes the published numbers never measured (halogen's tg rows are shallow-only).
|
||||
|
||||
## 2. Enumerate the undocumented levers
|
||||
The binary self-announces every env var it reads. The image often lacks `strings`:
|
||||
```
|
||||
podman run --rm --entrypoint /bin/sh <image> -c \
|
||||
'grep -aoE "<PREFIX>_[A-Z0-9_]+" /usr/local/bin/<engine> | sort -u'
|
||||
```
|
||||
Classify by kernel family (attention vs linear-state vs GEMV) against the suspected term. Only test levers with public precedent (e.g. issue #12 proved HALOGEN_FA_TQ) or an obvious name match; defaults are the publisher's measured winners — an A/B must show >=~15% to adopt.
|
||||
|
||||
## 3. Cache-assisted A/B (cheap reps)
|
||||
Boot each probe arm as its own container (env is read at startup) with the prompt cache ON (explicit MB >= one full-entry at ~70 KB/token): rep 1 pays the cold prefill, rep 2+ resumes (~2 s) — depth measurements then cost decode-time only. Publish the probe on host LOOPBACK (127.0.0.1:<scratch>), run the driver ON the host (python3 stdlib is enough), read the engine's own `timings` (prompt_ms/cache_n/predicted_ms/draft_n_accepted). Never benchmark beside a live engine on the same GPU.
|
||||
|
||||
## 4. Adopt or revert
|
||||
- Winner: env line into the serve env + test-seam update + docs, then `deploy.sh <target> --service <svc>` (restart restores everything; no unmask needed after stop-only).
|
||||
- Loser: `podman rm -f` probes, start the unit, revert nothing.
|
||||
- Either way: record the measured before/after in the runbook; report the depth behavior upstream if the publisher invites benchmark data — their bench gaps (depth, carve-topologies) are exactly what issues like #12 close.
|
||||
@@ -0,0 +1,13 @@
|
||||
---
|
||||
name: dankclaw-homepage-asset-verify
|
||||
description: "Add/verify homepage assets on Dankclaw: Caddy @data allowlist, deploy copy step, e2e panel census, live content-type check."
|
||||
---
|
||||
|
||||
Every new top-level homepage asset directory (e.g. `shared/homepage/img/`, `shared/homepage/sfx/`) needs FOUR coordinated changes, all in one commit:
|
||||
|
||||
1. **Caddy allowlist** — `dankclaw/Caddyfile` `@data path` matcher must include `/<dir>/*`. Without it the post-auth `@home` catch-all answers with `index.html` (`200 text/html` for a PNG/OGG — silent breakage).
|
||||
2. **Deploy copy** — `scripts/deploy-remote.sh` `deploy_homepage()` copies `shared/homepage/<dir>/` with the `cp -r ... /srv/homepage/` merge pattern beside the fonts/img/sfx blocks.
|
||||
3. **Test pins** — `tests/homepage_caddy.bats` grep-pins the new `@data` path; if the asset changes panel rendering, the e2e census in `tests/e2e/specs/10-homepage.spec.ts` must count the new render style (it derives expected names from services.json: `.panel .name` for normal cards, `.mi-name` for fused Model Information cards — a new fused component breaks the panel census otherwise, which is exactly how the 2026-10-08 Protoss deploy failed its e2e gate).
|
||||
4. **Deploy both services** — `deploy.sh dankclaw --service homepage` and `--service caddy` (Caddyfile change), then live-verify from a browser tab: `fetch('/<dir>/<file>')` must return `200` with the correct `content-type` (`image/jpeg`, `audio/ogg`), not `text/html`.
|
||||
|
||||
Checker note: `status.json` names follow `services.json` automatically (the deployed checker re-reads it via the homepage deploy's restart of homepage-status.timer), so renaming cards needs no checker change.
|
||||
@@ -0,0 +1,21 @@
|
||||
---
|
||||
name: defuse-shell-rc-secrets
|
||||
description: "Find literal API keys/tokens in shell rc files and replace them with 600-perm secret files + loader lines, syncing the canonical tool token file."
|
||||
---
|
||||
|
||||
# Defuse shell-rc secrets
|
||||
|
||||
Symptom: `export KEY=...` literal in `.zshrc`/`.bashrc` — world-readable in every process env, one `git add -A` from a dotfiles repo.
|
||||
|
||||
## Procedure (repeat per host)
|
||||
1. Sweep: `grep -n "export.*KEY\|export.*TOKEN" ~/.zshrc ~/.zprofile ~/.profile ~/.bashrc 2>/dev/null` (run per host; when batching hosts in one ssh compound command, label which output came from which host — misattribution wastes turns).
|
||||
2. Extract WITHOUT echoing: `mkdir -p ~/.config/secrets && chmod 700` it, then `sed -n 's/^export NAME="\(.*\)"$/\1/p' ~/.zshrc > ~/.config/secrets/<name>` (quote variant and unquoted variant are different sed patterns — verify byte counts with `wc -c`, a 0-byte file means the pattern missed).
|
||||
3. `chmod 600` every secret file; sync any canonical tool file the ecosystem expects (e.g. `~/.cache/huggingface/token`).
|
||||
4. Replace the rc literal (hashline `PUT N.=N:` range — never echo the secret in the op): `[ -f ~/.config/secrets/<name> ] && export NAME="$(cat ~/.config/secrets/<name>)"`.
|
||||
5. Verify: `zsh -n ~/.zshrc` (or `bash -n`); keep a dated `.bak` of the rc on remote hosts; shred any /tmp transfer copies immediately (`shred -u`).
|
||||
|
||||
## Laws
|
||||
- A token pasted into chat lands in transcripts and model context — treat chat as a leak channel; give the user the `read -rs` snippet instead.
|
||||
- scp of a secret FILE (encrypted channel, no argv) is fine; the token is never a command-line argument.
|
||||
- After replacing a legacy HF token, probe scope non-destructively: `HfApi().whoami()` (legacy read tokens show `type: access_token` without canWrite; the decisive probe is `create_repo(exist_ok=True)` on the real target repo).
|
||||
- Verify extraction byte-count before deleting/trusting; a successful-looking sed can write an empty file.
|
||||
@@ -0,0 +1,46 @@
|
||||
---
|
||||
name: detached-fleet-job-launch
|
||||
description: Launch a long-running remote job on Dank fleet hosts so it survives ssh/logind teardown
|
||||
---
|
||||
|
||||
# Detached long-running jobs on Dank fleet hosts
|
||||
|
||||
## Law
|
||||
`ssh host 'setsid nohup CMD & '` is NOT detached. systemd-logind deactivates
|
||||
`session-N.scope` when the ssh command returns and SIGKILLs the whole cgroup:
|
||||
no journal entry, no OOM trace, python block-buffers stdout so the log stays
|
||||
0 bytes. setsid/nohup do not escape the cgroup.
|
||||
|
||||
## Canonical launch
|
||||
```
|
||||
ssh host 'export XDG_RUNTIME_DIR=/run/user/$(id -u); \
|
||||
systemd-run --user --collect --unit=NAME \
|
||||
-p SendSIGKILL=no \
|
||||
-p StandardOutput=append:/home/dank/logs/NAME.log \
|
||||
-p StandardError=append:/home/dank/logs/NAME.log \
|
||||
sh -c "CMD -u; echo \$? > /home/dank/logs/NAME.rc"'
|
||||
```
|
||||
- transient user unit = outside every session scope; survives ssh teardown
|
||||
(requires `Linger=yes`, true for `dank` on both hosts)
|
||||
- `echo $? > …rc` inside the unit = completion sentinel the unit itself writes
|
||||
- `-u` / unbuffered output so the log is live
|
||||
- verify launch in the same ssh call: `systemctl --user is-active NAME.service`
|
||||
|
||||
## Watcher (one async job, never poll)
|
||||
```
|
||||
ssh host 'while [ ! -f /home/dank/logs/NAME.rc ]; do sleep 60; done; \
|
||||
echo RC=$(cat /home/dank/logs/NAME.rc); tail -25 /home/dank/logs/NAME.log'
|
||||
```
|
||||
Run it async with timeout 0; the result is injected on completion.
|
||||
|
||||
## Forensics when a detached job died
|
||||
1. `ps -p PID` — gone.
|
||||
2. Log 0 bytes → buffered stdout lost at SIGKILL.
|
||||
3. `journalctl --since … | grep -E "session-[0-9]+.*(Deactivated|Removed session)"` —
|
||||
timestamps matching the death window confirm the scope kill (vs OOM, which
|
||||
leaves kernel lines).
|
||||
|
||||
## Notes
|
||||
- zsh remote shells: `echo ===X===` fails (`=cmd` expansion); quote sentinels.
|
||||
- The HF upload of 2026-10-08 (113.7 GiB, hf_xet) died this way ~2 min in;
|
||||
relaunch as above succeeded.
|
||||
@@ -0,0 +1,27 @@
|
||||
---
|
||||
name: faillock-sudo-lockout-recovery
|
||||
description: "Diagnose and recover from a pam_faillock sudo lockout on Frank's Arch workstations, typically caused by an agent's non-interactive sudo retries"
|
||||
---
|
||||
|
||||
# Faillock Sudo Lockout Recovery
|
||||
|
||||
## Symptom
|
||||
User reports "sudo password rejected / changed"; journal shows
|
||||
`pam_faillock(sudo:auth): User <u> is temporarily locked out due to 3 consecutive failed login attempts`.
|
||||
|
||||
## Diagnose (read-only, one round)
|
||||
1. `stat -c '%n mtime=%y ctime=%z' /etc/shadow` — unchanged mtime = password was NEVER changed.
|
||||
2. `journalctl --since '<window>' | grep 'sudo\['` — lines with `PWD=` / `COMMAND=` name the spammer (agent harness loops show `PWD=/home/.../.omp`, `COMMAND=...yay/pacman...`, no `TTY=`).
|
||||
3. `faillock --user <user>` — shows the valid failure entries.
|
||||
|
||||
## Root cause law
|
||||
In the omp harness sudo has NO TTY: each `sudo <cmd>` = 3 pam_unix failures. Arch pam_faillock defaults deny=3 with ~600s unlock_time, and each retry re-arms it — the user's real interactive sudo is locked while the loop runs.
|
||||
|
||||
## Recover
|
||||
1. Kill the loop: `ps -eo pid,ppid,etime,cmd | grep -E 'sudo|yay|makepkg'` → kill offending PIDs; stop/kill the responsible agent session.
|
||||
2. `faillock --user "$USER" --reset` — works unprivileged for your own user (dir /run/faillock/<user> is 0700 owner).
|
||||
3. User verifies with `sudo -v` in a real terminal.
|
||||
|
||||
## Prevent
|
||||
- Agents NEVER run sudo/yay/pacman/pkexec; no SUDO_ASKPASS/expect workarounds.
|
||||
- Root-requiring steps are handed to the user as exact copy-paste commands in the final report.
|
||||
@@ -0,0 +1,19 @@
|
||||
---
|
||||
name: fandom-wiki-media-fetch
|
||||
description: Fetch audio/image media from a Fandom wiki via its MediaWiki API when the wiki HTML pages 403 curl or hang headless browsers.
|
||||
---
|
||||
|
||||
Goal: get direct media URLs (and files) from a starcraft.fandom.com-style Fandom wiki.
|
||||
|
||||
Why this route: Fandom HTML pages return a ~5 KB JS shell to `curl` (403 via `read ...:raw`) and time out headless Chromium (`browser.open`/`goto` >30 s on quotation megapages). The MediaWiki API is plain JSON and answers plain curl.
|
||||
|
||||
Steps:
|
||||
1. Find the content page's wikitext:
|
||||
`curl -sL -A "Mozilla/5.0" "https://<wiki>.fandom.com/api.php?action=parse&page=<Page_Name>&prop=wikitext&format=json&formatversion=2"`
|
||||
Top-level pages are often tabber/transclusion shells — follow the `{{:Subpage}}` / `{{TemplateName}}` references with the same call (e.g. `StarCraft_unit_quotations` → `/Protoss` → `Template:QuotesSC1Zealot`).
|
||||
2. Extract media names from the wikitext. StarCraft quotes use `{{sm2|File.ogg|displayed quote}}` inside `{{UnitQuoteBox|select=...}}` fields — match the quote text to pick the exact file.
|
||||
3. Resolve direct CDN URLs (works for images too):
|
||||
`curl -sL -A "Mozilla/5.0" ".../api.php?action=query&prop=imageinfo&iiprop=url|size&format=json&formatversion=2&titles=File:A.ogg|File:B.jpg"`
|
||||
→ `query.pages[].imageinfo[0].url` = `https://static.wikia.nocookie.net/...` — downloadable with plain `curl -sfL`.
|
||||
4. To discover art on a unit page: `action=query&prop=images&imlimit=100&titles=<Page>`, filter titles (icon/art/portrait/Head/Render), then resolve the keepers via imageinfo. Note splash/head art naming: `File:Zealot SC2 Art1.jpg`, `File:Archon SC2 Head1.jpg`; some obvious guesses (SC1-era units' SC2 splash) do not exist — verify with imageinfo, absent pages return `missing` with no `imageinfo`.
|
||||
5. Verify downloads (`file sfx/*.ogg` → `Ogg data, Vorbis audio`) — a 403 body saves as a success with `-o` and no `-f`.
|
||||
@@ -0,0 +1,22 @@
|
||||
---
|
||||
name: frank-primary-decode-tuning
|
||||
description: "Tune frank-primary (radiance TP=2) decode throughput: benchmark harness usage, flag levers, known optima and boot-failure walls."
|
||||
---
|
||||
|
||||
# frank-primary decode tuning
|
||||
|
||||
Harness: `bash autoresearch.sh` (dev box) scp's `scripts/frank-primary.env` + `scripts/autoresearch-frank-primary.sh` to dankserver, stops prod (`systemctl stop frank-primary`, NEVER mask — relinked unit), boots scratch container `arbench` on :7770 (fresh prefix-cache tier at `$STATE_DIR/kvcache/arbench` for cold prefill), polls health 720s, sends ONE greedy completion (fixed 140k-char repo-corpus prompt -> 43,627 tokens, 256-token decode), parses `timings` into METRIC lines, tears down, restores prod via EXIT trap. ~4 min/run; decode noise ~0.3%.
|
||||
|
||||
Metric: `METRIC decode_tps` = `timings.predicted_per_second` at 43.6k depth.
|
||||
|
||||
## Optima (measured 2026-10-06, keep these)
|
||||
- `FRANK_PRIMARY_SPEC_TOKENS=2` (unimodal: 1->31.89, 2->32.40, 3->31.72, 4->24.57 t/s; acceptance collapses past 3)
|
||||
- `FRANK_PRIMARY_MAX_SEQS=3` with `MAX_MODEL_LEN=200000` (600k KV; +4.2% total; 4 slots regresses)
|
||||
- `FRANK_PRIMARY_EXPERT_VS_CACHE=0.82` bracket-confirmed (0.78 fails boot, 0.84 -8.9%, 0.87 collapse)
|
||||
- `TP_WIRE=wht6`, `MAX_BATCHED_TOKENS=2048` — REQUIRED at seqs=3: bf16 wire and MNBT 4096 both fail boot (VRAM)
|
||||
|
||||
## Non-levers (decode-neutral, don't spend runs)
|
||||
GPU_HEADROOM_MIB (48 = neutral), HOST_POOL_MIB 49152 (neutral), MAX_MODEL_LEN any change (regresses: bigger KV steals resident weights), prefix-cache tiers (prefill-side).
|
||||
|
||||
## Rules
|
||||
One flag per iteration; keep only > +1% decode over best; a boot failure is `crash` (trap restores prod). Constraint floor: MAX_SEQS*MAX_MODEL_LEN >= 400000.
|
||||
@@ -0,0 +1,68 @@
|
||||
---
|
||||
name: git-repo-secret-audit
|
||||
description: "Audit a git repo for leaked secrets across history, unreachable objects, LFS payloads, git config and host configs — with triage of every hit."
|
||||
---
|
||||
|
||||
# Git repo secret audit (history + unreachable objects, not just the worktree)
|
||||
|
||||
Worktree greps prove nothing: the leak you care about is in a commit that was amended
|
||||
away, in a branch nobody checks out, in an LFS payload, or in `.git/config`.
|
||||
|
||||
## 0. Build the known-value set from the live secrets, never from literals
|
||||
Read the values from the secret store at runtime so no key enters the transcript:
|
||||
```python
|
||||
known = {}
|
||||
for p in (home/".config/secrets").iterdir():
|
||||
t = p.read_text().strip()
|
||||
vals = {t}
|
||||
m = re.search(r":([^:@/]+)@", t) # credential-store URL form
|
||||
if m: vals.add(m.group(1))
|
||||
known[p.name] = {v for v in vals if len(v) >= 12}
|
||||
```
|
||||
Report ONLY `value[:4] + len + sha256[:10]`. Never print a value, and never quote a
|
||||
secret-shaped literal in a doc/commit message — the pre-commit gate will (correctly) block it.
|
||||
|
||||
## 1. Shapes to match (superset of the Dank gate)
|
||||
`sk-[A-Za-z0-9]{18,}`, `sk-or-v1-`, `sk-ant-`, `hf_[A-Za-z0-9]{30,}`, `gh[pousr]_[A-Za-z0-9]{30,}`,
|
||||
`github_pat_`, `glpat-`, `xox[baprs]-`, `(AKIA|ASIA)[0-9A-Z]{16}`, `npm_`, `pypi-`, `AIza[0-9A-Za-z_-]{30,}`,
|
||||
`(sk|pk)_(live|test)_`, Telegram `\d{8,10}:AA`, JWT `eyJ…\.eyJ…`, `-----BEGIN [A-Z ]*PRIVATE KEY-----`,
|
||||
`AGE-SECRET-KEY-1`, `ssh-(ed25519|rsa)\s+AAAA`, `scheme://user:pass@host` basic-auth URLs,
|
||||
`Bearer <24+>`, and the noisy `(password|secret|api_key|token)\s*[:=]\s*<12+>` assignment shape.
|
||||
|
||||
## 2. Scope, in this order
|
||||
```bash
|
||||
git rev-list --all | wc -l # revs incl. every branch + tag
|
||||
git ls-tree -r --format='%(objectname) %(path)' <rev> # sha -> path, per rev
|
||||
git cat-file --batch # stream reachable blob content
|
||||
git cat-file --batch-all-objects --batch-check # EVERY object, incl. dangling
|
||||
git rev-list --objects --all # the reachable set; difference = dangling
|
||||
git log --all --oneline --find-object=<sha> # empty => never committed on any ref
|
||||
git status --porcelain -uall --ignored=matching # !! lines: on-disk ignored files
|
||||
git config --list --show-origin # local/global/system credentials
|
||||
git ls-files | grep -E '(auth\.json|trust\.json|\.env|\.pem|\.key|identity)' # quarantine
|
||||
```
|
||||
LFS: scan `.git/lfs/objects/**` as BYTES (`decode("latin-1")`) — a screenshot is a classic
|
||||
secret carrier and the tracked tree holds only pointers.
|
||||
|
||||
## 3. Entropy layer (catches issuer-specific keys you did not list)
|
||||
Tokens `[A-Za-z0-9+/_-]{32,}` with Shannon entropy >= 4.3 containing both digits and letters.
|
||||
Flag every one whose value is in `known` as a LIVE leak; list the rest for triage.
|
||||
|
||||
## 4. Triage — the shapes that always fire and are fine
|
||||
- `api_key="sk-no-key-required"` and other doc placeholders
|
||||
- prose: `Set password: user.set_password(...)`, `authenticate(user, password)`
|
||||
- 40-hex hits that are git object ids / packfile names (`pack-<40hex>.pack`)
|
||||
- memory/KB records that *describe* a secret operation
|
||||
A hit is a leak only when the value matches a live secret, decodes to a real credential, or
|
||||
sits in a field a program reads.
|
||||
|
||||
## 5. Adjacent host scan (the audit's real yield)
|
||||
Check file modes of the secret dir (`0600`, dir `0700`) — world-readable secret files are the
|
||||
most common real finding. Then grep the live values across shell rc files, `~/.gitconfig`,
|
||||
`~/.npmrc`, `~/.netrc`, `~/.curlrc`, `~/.docker/config.json`, `~/.cache/huggingface/token`,
|
||||
`~/.aws/credentials`, `~/.ssh/config` and report filenames only.
|
||||
|
||||
## 6. Report
|
||||
Per layer: scope numbers + verdict. Every hit: path:line, redacted value, verdict
|
||||
(leak / benign + why). Name the gate's coverage gap if the project's own hook misses shapes
|
||||
you matched. Fix modes/chmod on the spot; changes to a security hook are the operator's call.
|
||||
@@ -0,0 +1,52 @@
|
||||
---
|
||||
name: gitea-token-out-of-remote-url
|
||||
description: "Move a Gitea/HTTP git token out of remote.origin.url into a credential store file, and verify auth works with prompts disabled."
|
||||
---
|
||||
|
||||
# Move a git token out of the remote URL
|
||||
|
||||
A remote URL like `http://user:<token>@host:port/org/repo.git` puts the credential in
|
||||
`.git/config` (copied by every clone), prints it in `git remote -v`, and git-lfs echoes the
|
||||
whole URL in its `Locking support detected on remote` warning on every push.
|
||||
|
||||
## 1. Capture the token from the existing remote (never paste it into a message)
|
||||
```bash
|
||||
url=$(git remote get-url origin)
|
||||
user=$(printf '%s' "$url" | sed -E 's#^http://([^:]+):[^@]+@.*$#\1#')
|
||||
tok=$(printf '%s' "$url" | sed -E 's#^http://[^:]+:([^@]+)@.*$#\1#')
|
||||
```
|
||||
|
||||
## 2. Write the credential file (URL form, port included)
|
||||
```bash
|
||||
umask 077
|
||||
printf 'http://%s:%s@HOST:PORT\n' "$user" "$tok" > ~/.config/secrets/gitea
|
||||
```
|
||||
Git 2.56 `credential-store` does NOT parse the `protocol=`/`host=`/`password=` block shape
|
||||
that `git credential fill` accepts on stdin: a block-format file gives a SILENT empty get,
|
||||
then git falls through to askpass (`could not read Username`). One URL per line, port included.
|
||||
|
||||
## 3. Point git at it, strip the URL, silence the LFS echo
|
||||
```bash
|
||||
git config --global credential.helper 'store --file ~/.config/secrets/gitea'
|
||||
git remote set-url origin http://HOST:PORT/org/repo.git
|
||||
git config lfs.http://HOST:PORT/org/repo.git/info/lfs.locksverify true
|
||||
```
|
||||
|
||||
## 4. Verify with prompts disabled (a hang is a failed credential)
|
||||
```bash
|
||||
printf 'protocol=http\nhost=HOST:PORT\n\n' | git credential-store --file ~/.config/secrets/gitea get
|
||||
GIT_TERMINAL_PROMPT=0 git ls-remote --heads origin main # read auth
|
||||
GIT_TERMINAL_PROMPT=0 git push --dry-run origin main # write auth
|
||||
git remote get-url origin # must contain no '@'
|
||||
grep -l "$tok" ~/.gitconfig */.git/config ~/.zshrc # must find nothing
|
||||
```
|
||||
A successful push over the bare URL is the end-to-end proof: the credential came from the file.
|
||||
|
||||
## 5. Sweep the leak
|
||||
Same shape on every clone on the host — loop over `find ~ -maxdepth 4 -path '*/.git/config'`,
|
||||
read `remote.origin.url`, scrub any URL containing `@`. One credential file serves them all
|
||||
via the global helper.
|
||||
|
||||
## Rotate
|
||||
Rewrite only the credential file, then delete the old token in the Gitea UI
|
||||
(Settings → Applications). No remote URL, rc file, or repository changes.
|
||||
@@ -0,0 +1,26 @@
|
||||
---
|
||||
name: hf-checkpoint-profile
|
||||
description: "Profile a Hugging Face safetensors checkpoint (sizes, dtypes, tensor-name space, gated/license) with compact HTTP-range reads — safe for subagents on small context slots."
|
||||
---
|
||||
|
||||
# Profile an HF safetensors checkpoint cheaply
|
||||
|
||||
Goal: inventory a repo (files/bytes, dtype histogram, tensor-name patterns, per-group GiB, config keys, license/access) without pulling gigabytes and without overflowing a subagent's context slot.
|
||||
|
||||
## Endpoints (all public, no token needed even for gated models)
|
||||
- `https://huggingface.co/api/models/<id>?blobs=true` → `siblings` (name+size), `gated`, `private`, `downloads`, `safetensors` metadata, `tags`.
|
||||
- `https://huggingface.co/api/models/<id>` for tags/base model; `https://huggingface.co/api/models?author=<x>&limit=100` to find sibling repos (canonical id check: cards often link both `X-1.5-Y` and `X-Y`).
|
||||
- `https://huggingface.co/api/models/<id>/resolve/main/<file>` for `config.json`, `model.safetensors.index.json`, `chat_template.jinja`, `LICENSE`, `generation_config.json`; `/raw/main/<file>` for text.
|
||||
|
||||
## dtype + per-group bytes without downloading shards
|
||||
Parse the safetensors header over HTTP Range: first 8 bytes = little-endian u64 header length, then that many bytes of JSON (`{name: {dtype, shape, data_offsets}}`). Fetch every shard header with a thread pool (131 shards ≈ seconds), then bytes = prod(shape) × dtype size. Collapse names with `re.sub(r"\.\d+\.", ".N.", k)` (and `\.shard_\d+` → `.shard_N`) to get the pattern space; group-sum by prefix to see where the mass sits (experts / ngram-PLE / MTP / vision / trunk).
|
||||
|
||||
Compare the key set against the base model's index to prove "full-weight finetune, identical names" (a zero-diff key set means per-tensor recipes written for the base apply verbatim, and it reveals fused vs per-expert layout, PLE/ngram shards, MTP head, vision tower, absence of `quantization_config`).
|
||||
|
||||
## Guardrails
|
||||
- Run python scripts from a clean directory: a stray `/tmp/inspect.py` shadows stdlib `inspect` and kills `concurrent.futures` imports.
|
||||
- Never put an HF token on a command line (`/proc/<pid>/cmdline` is world-readable); use `curl -K <mode-600 file>`.
|
||||
- `curl -f` / `--fail-with-body` and verify byte counts — a 401/403 writes the error body as the "artifact" and exits 0.
|
||||
|
||||
## Subagent context budget
|
||||
Subagents on a 131k-token slot die from `exceed_context_size_error` when told to read whole doc pages. Give grep-only instructions (`| head -40`, `-o`, `-c`), demand a line cap ("report in ≤250 lines, no verbatim dumps"), and prefer `git clone --depth 5` + local grep to web-page reads. A peer scout that overflows returns exit 1 with the 400 in the harness log and NO structured output — assume nothing was learned and redo the fact inline.
|
||||
@@ -0,0 +1,40 @@
|
||||
---
|
||||
name: hf-large-artifact-publication
|
||||
description: "Publish a large quantized artifact + model card (and an optional anonymized corpus dataset) to HF Hub from dankserver: staging set, HF-style card spine, upload gates, transient-unit launcher, sentinel watching, card-metadata and scrub laws."
|
||||
---
|
||||
|
||||
# HF large-artifact publication (dankserver)
|
||||
|
||||
Publish a big model artifact (100+ GiB) + card, and optionally an anonymized corpus dataset, from dankserver under `frank8465`.
|
||||
|
||||
## Staging
|
||||
- Dev box `~/Projects/hf-swift15-upload/` (author card + scripts here), `scp -r` synced to `/home/dank/hf-swift15-upload/` on dankserver. Keep staging + logs + `.rc` sentinels after publication as receipts.
|
||||
- Payload set: artifact, `README.md` card, `scripts/` (recipe/drivers/probes), `artifacts/` (gate JSONs, per-position `.rows`, manifests, run logs).
|
||||
- Secret-scan the whole payload before arming: credential regexes (hf_/ghp_/pat, private keys, Bearer values, AKIA/sk-), LAN IPs, hostnames, domains, emails. Verdict each hit — variable *references* and upstream test fixtures are not leaks; check real values in context.
|
||||
|
||||
## Card law (HF validator)
|
||||
- Custom licenses MUST be `license: other` + `license_name:` + `license_link:` (mirror the base model's card trio — fetch `https://huggingface.co/api/models/<base>` and read `cardData`). A raw non-SPDX `license:` value raises `ValueError: Invalid metadata in README.md` inside `upload_file` — AFTER all local gates have run.
|
||||
- Card spine: what-the-format-is + engine pin digest, recipe table, quality gates (KL/top-1/PPL + per-position data shipped), proof hardware, exact `podman run` line, license chain, manifest table with the artifact's sha256 receipt, "In plain English:" block per section.
|
||||
- Quote the SHIPPED filename (build name), not a prettified one — a hub-side rename of a 100 GiB LFS file costs a full re-upload.
|
||||
|
||||
## Upload mechanics
|
||||
- venv `/home/dank/venvs/radcorpus` (huggingface_hub 1.28 + hf_xet); `HF_XET_HIGH_PERFORMANCE=1`; token from `~/.cache/huggingface/token` (see defuse-shell-rc-secrets pattern: `~/.config/secrets/<name>` 600 + rc loader, mirrored to the cache file). NEVER token in argv/chat.
|
||||
- Script gates, in order: sha256 of live artifact vs recorded digest (`gate 1 : MATCH`), `create_repo(..., private=False, exist_ok=True)`, small files first, artifact last, post-upload census — **compare remote sizes to local and FAIL the RC on any mismatch** (this caught a stale card mid-campaign).
|
||||
- hub 1.28 API traps: `upload_file`/`create_commit` kwargs are keyword-only; no `rename_file`, `hf_hub_url`, `HfApi.commit_api` (use `CommitOperationAdd`/`CommitOperationDelete` from `huggingface_hub` top-level if needed); `whoami` hides `canWrite` for legacy tokens — probe write scope by `create_repo(exist_ok=True)` on the real target.
|
||||
- zsh-over-ssh: never `echo ===` (triggers `=cmd` expansion); quote it.
|
||||
|
||||
## Launcher + watchers
|
||||
`setsid nohup` over ssh dies at session teardown. Use a transient user unit:
|
||||
```
|
||||
systemd-run --user --collect --unit=hf-upload -p SendSIGKILL=no \
|
||||
-p StandardOutput=append:/home/dank/logs/hf-upload.log \
|
||||
-p StandardError=append:/home/dank/logs/hf-upload.log \
|
||||
sh -c "cd <staging> && <venv>/bin/python -u <uploader>.py --go; echo \$? > /home/dank/logs/hf-upload.rc"
|
||||
```
|
||||
Sentinel: background ssh loop polling the `.rc` file (60 s period, bounded), tails the log on appearance. Progress bars are `\r`-fed — `tr '\r' '\n' | tail` to read them.
|
||||
|
||||
## Corpus dataset variant
|
||||
- License audit first: read `urls.txt`/manifest, fetch each source card (`license` in cardData). Share-alike sources (CC BY-SA) → publish derived corpus CC BY-SA with attribution; model weights unaffected.
|
||||
- Fleet-harvested slices = agent-session transcripts: scrub with deterministic substring-class rules (NOT word-boundary: catches `dankclaw_default`, `FRANKBOOK_SVCS`, `franksdank/alloy` image namespace): LAN IPs→10.x, hostnames→server-a/b/client-c, domain→example.com (catch `\.`-escaped forms + bare brand), `/home/<user>`, emails→user@example.com. Residual probes run on OUTPUT; whitelist self-replacements + audited fixture noise.
|
||||
- Publish original hashes, disclose non-byte-identity, cross-link model card ↔ dataset card, verify via `raw/main/README.md` fetches.
|
||||
- Dataset repo needs dataset-write scope (fine-grained token; access tokens 403). `repo_type="dataset"` everywhere; 45 MB rides in ~60 s.
|
||||
@@ -0,0 +1,21 @@
|
||||
---
|
||||
name: homepage-smoke-verify-loop
|
||||
description: "Smoke-verify Dankclaw homepage changes locally before commit/deploy: isolated static tree + fake status + browser census, then deploy gate."
|
||||
---
|
||||
|
||||
# Homepage smoke-verify loop (serverAppConfig)
|
||||
|
||||
Verify `shared/homepage/` changes against a real browser BEFORE commit/deploy.
|
||||
|
||||
## Procedure
|
||||
1. `node --check shared/homepage/app.js` and `python3 -c "import json; json.load(open('shared/homepage/services.json'))"`.
|
||||
2. Isolated tree: `rm -rf /tmp/hp-smokeN && cp -r shared/homepage /tmp/hp-smokeN`.
|
||||
3. Fabricate runtime files in the tree:
|
||||
- `status.json`: `{"checked":"x","services":[{"name":<every services.json name>,"state":"up"}]}` (set one model `down` to exercise the red-pulse path).
|
||||
- `deployed.js`: `window.DEPLOYED_AT="smoke"`.
|
||||
4. Serve: `(cd /tmp/hp-smokeN && nohup python3 -m http.server 8099 >/dev/null 2>&1 & echo $! > /tmp/.hp-pid)`.
|
||||
- If reusing port 8099: `pkill -f "http.server 8099"` FIRST — a second server on a busy port fails silently and the OLD tree keeps serving.
|
||||
5. Browser census via eval tab on `http://127.0.0.1:8099/?v=<epoch>` — ALWAYS cache-bust with `?v=`; python http.server gets heuristically disk-cached by the browser. Assert: broken images (`.app-art,.mi-art` naturalWidth==0), computed styles (`getComputedStyle` for palette/fonts/clipPath), presence of new classes, `document.querySelectorAll('.docs section').map(s=>s.id)` for doc order, regex text census for banned words.
|
||||
6. Screenshot full page + `selector` shots; inspect visually.
|
||||
7. `./tests/run.sh` (expect only the known frank-primary MAX_SEQS failure), commit, `./scripts/deploy.sh dankclaw --service homepage` (runs Playwright e2e gate), then repeat the census against `https://www.franksdank.cloud/?v=<epoch>`.
|
||||
8. Cleanup: kill server, move `/tmp/hp-smokeN` to `~/.AGENTTRASH/YYYYMMDD-homepage-smoke/` (never `rm`).
|
||||
@@ -0,0 +1,15 @@
|
||||
---
|
||||
name: homepage-theme-smoke
|
||||
description: "Locally smoke-test the Dankclaw homepage theme (style.css/app.js/docs.html) in the managed headless browser, with the stale-server traps that fake the result."
|
||||
---
|
||||
|
||||
Goal: prove a homepage theme change on a throwaway tree before commit/deploy.
|
||||
|
||||
Protocol:
|
||||
1. Stage a fresh copy: `cp -r shared/homepage/. /tmp/hp-smoke-<rand>/`, drop `.AGENTTRASH` dirs, and write a fake `status.json` (`{"generated_at":"x","services":[{"name":...,"state":"up"}]}`) — app.js fetches services.json + status.json and renders NOTHING without both.
|
||||
2. Serve: `cd <tree> && nohup python3 -m http.server 8099 --bind 127.0.0.1 &`.
|
||||
3. STALE-LISTENER TRAP (fakes every census): a survivor server from an older wave keeps answering on 8099 — `ss -ltn | grep 8099` proves a listener exists, not WHICH tree it serves. After every (re)start verify identity: `for p in $(pgrep -f "http.server 8099"); do readlink /proc/$p/cwd; done` must print the NEW tree; then `curl -s localhost:8099/<file> | grep -c <new-marker>`. `kill` the stale pid — never trust a `||`-guarded restart. Also: `pkill -f "http.server 8099"` from a shell whose own command line contains that string kills the shell; kill by pid.
|
||||
4. Browser census: navigate with `?v=<epoch>` cache-bust (browser disk-caches http.server responses); `waitForSelector('.mi-card', ...)` before evaluating — rendering is async (fetch). Geometry checks: rect-gap to 1px for bottom-pinned chips; screenshot for visual proof. `screenshot({path})` may not write — use the returned omp-sshots artifact.
|
||||
5. In headless pages `new Audio()` is detached: prove voice playback via `tab.requests()` (206 on the /sfx file) + the `mi-speaking` class, not a DOM `<audio>` query.
|
||||
6. Live-domain check after deploy: the homepage is at `/`, the docs page ONLY at `/docs/` — `/docs.html` falls through the SPA fallback and serves the homepage shell. Anubis adds a challenge: `waitForSelector` with 30-45 s timeout.
|
||||
7. Deploy loop: `./tests/run.sh` → commit/push → `./scripts/deploy.sh dankclaw --service homepage` (runs the Playwright e2e gate) → live census with fresh `?v=`.
|
||||
@@ -0,0 +1,50 @@
|
||||
---
|
||||
name: llama-server-arch-verify
|
||||
description: Verify llama.cpp binary supports a model architecture without relying on binary grep
|
||||
---
|
||||
|
||||
# llama-server Arch Verification
|
||||
|
||||
## Problem
|
||||
`grep -ac <arch> llama-server` returns 0 for ALL architecture names (including ones the binary is actively serving). The arch strings are stored in a packed/constructed form that grep can't catch.
|
||||
|
||||
## Solution
|
||||
Verify arch support via the source registration table, not binary grep:
|
||||
|
||||
1. **Check the source registration** (`src/llama-arch.cpp` or `src/llmarch.h`):
|
||||
```bash
|
||||
grep -n "gemma4\|qwen3_5\|<arch>" src/llama-arch.cpp
|
||||
```
|
||||
If the arch is registered in the source at the commit the binary was built from, it's supported.
|
||||
|
||||
2. **Verify the binary is current**:
|
||||
```bash
|
||||
stat -c "%y" /path/to/llama-server
|
||||
git log -1 --format=%cd <commit>
|
||||
```
|
||||
The binary mtime should be after the commit date.
|
||||
|
||||
3. **Check for model-specific source files**:
|
||||
```bash
|
||||
ls src/models/<arch>.cpp src/models/<arch>-*.cpp 2>/dev/null
|
||||
```
|
||||
|
||||
4. **Check for spec-type support** (for MTP/draft models):
|
||||
```bash
|
||||
/path/to/llama-server --help 2>&1 | grep "spec-type\|model-draft"
|
||||
```
|
||||
|
||||
## KV Cache Types
|
||||
The build's allowed `--cache-type-k` / `--cache-type-v` values:
|
||||
`f32, f16, bf16, q8_0, q4_0, q4_1, iq4_nl, q5_0, q5_1`
|
||||
|
||||
**No true fp8 KV dtype** — `q8_0` is the "FP8 class" (1.0625 B/elem, 8.5 effective bits).
|
||||
|
||||
## Separate Draft Model (MTP assistant)
|
||||
For models where the MTP head is a SEPARATE file (not in-file like Swift):
|
||||
```bash
|
||||
--spec-type draft-mtp \
|
||||
--spec-draft-n-max 3 \
|
||||
--spec-draft-model /path/to/mtp-assistant.gguf
|
||||
```
|
||||
The `--spec-draft-model` flag (aka `-md`, `--model-draft`) loads the separate draft file.
|
||||
@@ -0,0 +1,28 @@
|
||||
---
|
||||
name: llamacpp-endpoint-model-swap
|
||||
description: "Swap the model on a dank-fleet llama.cpp endpoint (frank-secondary pattern): fetch, env edit, spec flags, deploy through git, rewrite the bats seam, verify VRAM with the engine's own numbers."
|
||||
---
|
||||
|
||||
# llama.cpp endpoint model swap (dank fleet)
|
||||
|
||||
Use when replacing the model/shape on a llama.cpp endpoint (e.g. frank-secondary on the RTX 3090). The engine's boot numbers are the verdict — do NOT commit to a feasibility verdict from head math; boot it and read `nvidia-smi`.
|
||||
|
||||
## Sequence
|
||||
|
||||
1. **Geometry first, from the loaded model's own log** (`--log-verbosity 4`, or the HF `config.json` under `text_config` for multimodal): layers, `n_head_kv` (an ARRAY — per-layer, e.g. gemma4 has 8 on sliding layers, 2 on full-attention), head_dim, `sliding_window_pattern`, `n_swa`, `attention_k_eq_v`. KV scales only with full-attention layers; sliding layers cost a fixed window-size ring. Single-value HF fields lie about per-layer heterogeneity.
|
||||
2. **Fit the KV rate law to the engine, not to arithmetic.** Run once with `--log-verbosity 4` and read the `llama_kv_cache` / `sched_reserve` buffer MiB lines. Known law for master 5fc4f3c: bytes/elem = (K bytes + V bytes) **x2** when the file declares `attention_k_eq_v` (6.125 at f16 K + q8_0 V, 2.125 at q8_0/q8_0) — k_eq_v is a compute-graph mapping and does NOT shrink the cache. Full cache = n_full x n_head_kv x head_dim x rate x ctx; SWA cache = n_swa x window x n_head_kv_swa x head_dim x rate. Both reproduce the printed MiB exactly; if your formula does not, your law is wrong, not the engine.
|
||||
3. **Probe the binary's real capabilities** before promising flags: `--help | grep -A2 cache-type-k` (this fleet's build has NO fp8; q8_0 is the "FP8 class"), `--help | grep -A2 verbosity` (`-lv, --log-verbosity N`; ship `VERBOSITY` as an env tunable + `--log-verbosity "$VERBOSITY"` so VRAM forensics land in the journal), `--help | grep model-draft` for separate drafters. NEVER `grep -c <archname>` on the packed binary to check arch support — it returns 0 for every arch; check the source tree at the build's commit.
|
||||
4. **Fetch with `-f` + byte-count check** into `/home/dank/models/<name>/`.
|
||||
5. **Edit the repo trio** (`scripts/serve-<svc>.sh`, `scripts/<svc>.env`, `~/.omp/agent/models.yml`): spec flags move as a SET (`SPEC_TYPE=none` strips `--spec-type` AND `--spec-draft-model`/`--spec-draft-*`); a drafter with its own arch file (gemma4-assistant) needs `SPEC_DRAFT_FILE`, and its draft KV is built against the TARGET's full window — price it from a boot pair. `models.yml` `contextWindow` = per-slot window (`CTX_SIZE/PARALLEL`); recompute `maxTokens` for the compaction clamp `min(maxTokens, ctx - estimate - 4096)`.
|
||||
6. **Rewrite the bats seam in the same commit** (`tests/<svc>_unit.bats`): payload path, slot geometry, KV pair, spec state, batch pair, verbosity, mmproj opt-out, reasoning flags — all argv-asserted against the stub engine; where the contract is a repo value (batch pair, verbosity) assert argv == env value so a retune moves one file.
|
||||
7. **Deploy through git** (no direct remote edits). If the remote tree is dirty: `git stash push <files>` on the host, `git pull --ff-only`, land the change as a commit. A pull that aborts leaves the host at the OLD revision while the service runs the NEW edits — verify `git log -1` on the host.
|
||||
8. **Verify the engine, not your math**: `/proc/<pid>/cmdline` (argv is the contract), `/v1/models` (`n_ctx`, `ftype`, `n_ctx_train`), `nvidia-smi --query-gpu=memory.used,memory.free`, `/health` 200, a vision probe (solid color >=256 px via `image_url`, expect the exact word), and one generation probe reading `timings.predicted_per_second`. Record the numbers in the env comment + runbook, then push and ff-pull the host.
|
||||
9. **Cutover AGENTS.md in the same commit**: fleet-topology table row, the `### <svc>` section, the `Model Config` contextWindow/vision note, the runbook index ACTIVE row, a RETIRED row for the superseded era. The new docs page must be linked from the TOP-LEVEL `serverAppConfigs-docs/index.md`.
|
||||
|
||||
## Gotchas
|
||||
|
||||
- A systemd restart reporting `inactive` in the first seconds is usually model-load time; check `journalctl -u <svc>` for `model done`/`server is listening` before declaring failure.
|
||||
- Rollback = the previous payload stays on disk; one `MODEL_FILE` + shape edit back.
|
||||
- A thinking model can consume the whole token budget in `reasoning_content` (empty `content`, `finish_reason: length`) — probe with a budget that clears thinking.
|
||||
- **Bats vacuity trap**: `serve_argv` REWRITES `$STAGE/test.env` from the shipped env, so a test that mutates the file (`sed -i`, append) then calls `serve_argv` asserts on regenerated content — a tautology. Tests that mutate must exec via a `run_stage` helper running the file as-is. Symptom: the assertion fails only after you fix the vacuity, and the env dump shows the mutation gone.
|
||||
- `env NAME=value cmd` needs valid identifiers; a flag-like token (`nvidia-modprobe=true`) is only legal as an `env` ARGUMENT, not as a shell assignment prefix.
|
||||
@@ -0,0 +1,27 @@
|
||||
---
|
||||
name: llamacpp-serve-unit-boot-loop
|
||||
description: "Diagnose a Dank-fleet LLM systemd unit that boot-loops: classify 203/EXEC vs model-load OOM vs payload gate, using journal signatures and nvidia-smi/rocm-smi."
|
||||
---
|
||||
|
||||
# llama.cpp / fleet serve unit boot loop
|
||||
|
||||
A fleet LLM unit (`frank-large`, `frank-secondary`, `frank-primary`) restarts every `RestartSec`. Read the journal FIRST, act second:
|
||||
|
||||
```
|
||||
ssh <host> 'sudo systemctl status <svc> --no-pager | head -12; sudo journalctl -u <svc> --since "-6 min" -o cat --no-pager | tail -40'
|
||||
```
|
||||
|
||||
## Classify by signature
|
||||
|
||||
| Signature | Cause | Fix |
|
||||
|---|---|---|
|
||||
| `status=203/EXEC` + `Failed at step EXEC spawning` + `Mem peak: ~1.9M` + GPU flat at 1 MiB | the `ExecStart` script never ran: missing shebang, lost exec bit, or path moved by the deploy (git pull mid-restart) | `head -1 <script>` must be `#!/usr/bin/env bash`; `git ls-files -s <script>` must show `100755`; commit the fix, redeploy |
|
||||
| `status=203/EXEC` right after a deploy, then self-heals | git replaced the file mid-checkout while the restart loop ran | ignore if the newest attempts succeed; verify `podman ps --format '{{.CreatedAt}}'` identity |
|
||||
| Engine prints weights loading, then dies | VRAM/disk budget, engine flags | read the engine's own ledger (verbosity 4) for KV/compute-buffer MiB; price from two boots, never a formula |
|
||||
| Health 503, no error, minutes of silence | normal cold load (big PLE/mmap) | do not kill; the listener binds after the full load |
|
||||
|
||||
## Rules
|
||||
- Judge the newest container/attempt only: `podman ps -a --format '{{.ID}} {{.CreatedAt}}'`, `journalctl --since`; a mid-stop container still serves old logs.
|
||||
- `systemctl is-active` is `failed` after every intentional stop on foreground-podman units (exit 143) — cosmetic.
|
||||
- The unit's `Description` is the last-deployed text, not the repo's. A stale Description is not an incident; the next `relink_unit` fixes it — never restart a serving endpoint for cosmetics.
|
||||
- Fixing the script = commit + `./scripts/deploy.sh <host> --service <svc>`; never edit on the host (a direct `scp` leaves drift that blocks the next deploy's ff-only pull with "local changes would be overwritten"; repair with `git checkout -- <file>`).
|
||||
@@ -0,0 +1,27 @@
|
||||
---
|
||||
name: llm-bench-noise-discipline
|
||||
description: "Noise-floor discipline for single-run LLM endpoint A/B benchmarks: the ±3% law, replication-before-keeping, and bracket-confirmation of tuned values."
|
||||
---
|
||||
|
||||
# LLM Bench Noise Discipline
|
||||
|
||||
Single-run A/B benchmarks against an LLM endpoint's own timings API have a much larger run-to-run noise floor than the metric's instantaneous precision suggests. On frank-primary (radiance, 256-token greedy decode at 43.6k-token prompt depth), three runs of the IDENTICAL config measured 33.06 / 32.03 / 30.90 t/s — a ~3% spread.
|
||||
|
||||
## The ±3% law
|
||||
- Treat any single-run delta under ~3% as noise, regardless of the harness's reported confidence multiplier.
|
||||
- Before believing a "keep", run the same config a second time (no edit, pure replication). If the replication drops below the noise band, flag the lucky draw as suspect (`flag_runs`) so best-metric math stops trusting it.
|
||||
- Noise sources: TP all-reduce wire loss (wht6), page-cache warm/cold of the 51 GiB n-gram table, container boot ordering, expert-pool hit-rate warm-up.
|
||||
|
||||
## Bracket before trusting an optimum
|
||||
A "tuned" value is only confirmed when BOTH neighboring steps are measured: one side may crash (boot wall) and the other may regress. Example at seqs=3: EXPERT_VS_CACHE 0.78 crash / 0.82 optimum / 0.84 = -9% (0.87 collapses to 0% per runbook); MNBT 4096 crash at seqs=3 while the seqs=2 A/B showed it bootable.
|
||||
|
||||
## Boot failures are data, not waste
|
||||
A scratch arm that never reaches /health (720s timeout) maps the viable envelope. Three crashes at seqs=3 (0.78 weights-share, 4096-token chunks, bf16 wire) prove the production trio wht6/2048/0.82 is exactly what makes the 3-slot shape fit.
|
||||
|
||||
## Outside-noise signals found at this endpoint
|
||||
- SPEC_TOKENS=4: -23% decode, draft acceptance 0.831->0.497 (the 4th draft token is rejected too often to pay its verification).
|
||||
- SPEC_TOKENS=0: -6.8%; MTP drafting always pays.
|
||||
- KV shape seqs=4/maxlen=100k (same 400k total): -11%; slot overhead is not paid by freed VRAM.
|
||||
|
||||
## Harness pattern that worked
|
||||
stop prod -> boot scratch container on the SAME two GPUs with candidate argv -> poll /health -> ONE self-contained greedy request built from a fixed-seed repo corpus (deterministic prompt_n) -> parse timings -> tear down -> restore prod via trap (never mask a relinked unit; stop only). Fresh prefix-cache dir per run so the cold prefill is cold.
|
||||
@@ -0,0 +1,27 @@
|
||||
---
|
||||
name: noctalia-bar-layout-edit
|
||||
description: "Edit/verify Noctalia v5 bar widget layout (start/center/end, taskbar workspace grouping) on Frank's symlinked dotfiles workstations"
|
||||
---
|
||||
|
||||
# Noctalia v5 bar layout edit
|
||||
|
||||
Bar widget layout on Frank's noctalia v5 workstations is GUI-override state, not config.toml.
|
||||
|
||||
## Where
|
||||
- Layout file: `~/.local/state/noctalia/settings.toml` → `[bar.<name>]` keys `start` / `center` / `end` (arrays of widget ids; `[widget.<id>]` subtables hold per-widget settings).
|
||||
- On Frank's boxes this file is a **symlink into the command-center repo** (e.g. `~/.omp/dotfiles/noctalia/.local/state/noctalia/settings.toml`) — edit the repo file, not the link path (atomic-rename writes through the link path would replace the symlink).
|
||||
- Layer order: built-in defaults < `~/.config/noctalia/config.toml` (+`[include]`) < `settings.toml` (wins). Noctalia writes through symlinks and keeps them.
|
||||
- Defaults when a key is absent: `start = ["launcher","wallpaper","workspaces"]`, `center = ["clock"]`.
|
||||
|
||||
## Apply/verify loop (no restart, no reload IPC)
|
||||
1. Edit the repo `settings.toml`.
|
||||
2. `noctalia config validate` (checks merged layers; also `noctalia config export [full] | yq -p toml` to inspect effective values).
|
||||
3. Both layers are file-watched → hot-reload. Prove visually: `noctalia msg screenshot-fullscreen` saves a PNG to `~/Pictures`; read the image and inspect the bar.
|
||||
|
||||
## Taskbar recipes
|
||||
- Move window indicator beside workspaces: put `"taskbar"` in `start` after `"workspaces"`.
|
||||
- Combine workspaces+windows into ONE widget: `[widget.taskbar] group_by_workspace = true` — per-workspace capsules with disc labels (focused/occupied/empty/urgent colors); `workspace_group_content = "icons"|"count"|"dots"` (count/dots: left-click activates the workspace); `hide_empty_workspaces`, `show_workspace_label`, `minimal`, `workspace_label_placement = corner|centered|inside`. Requires compositor per-window workspace assignment (niri: yes; LabWC: no → stays flat).
|
||||
- The `workspaces` widget itself has no window display; styles are regular/minimal/focus_hint only.
|
||||
|
||||
## Law
|
||||
Full widget/setting tables: https://docs.noctalia.dev/noctalia/bar/ and .../bar/widgets/<name>/ . Commit bar changes atomically to the command-center repo (checkout is the deployment).
|
||||
@@ -0,0 +1,24 @@
|
||||
---
|
||||
name: omp-harness-window-fit-audit
|
||||
description: "Audit an omp harness model entry against an engine's prompt+max_tokens fit check; explains and fixes 400 \"max_tokens does not fit\" at 60-85 % of the declared window."
|
||||
---
|
||||
|
||||
Use when a local endpoint returns 400 `max_tokens N does not fit: prompt is P tokens and the context is C, leaving room for R` on omp, typically at 65-85 % of the declared window.
|
||||
|
||||
## The three numbers that must agree
|
||||
1. Engine hard ceiling: prompt_tokens + max_tokens ≤ window (halogen `HALOGEN_CTX`/slot, llama.cpp per-slot `ctx_size/parallel`, radiance `--max-model-len`). A 400, never truncation.
|
||||
2. Harness clamp: `max_tokens = min(maxTokens, contextWindow − est − 4096)`, est = chars/4 over the conversation+tools. The dump in `~/.omp/logs/http-400-requests/*.json` proves it: est = contextWindow − 4096 − max_completion_tokens.
|
||||
3. Compaction fires only at `compaction.thresholdPercent` × contextWindow (config.yml; ignore stale reserveTokens/70 % comments).
|
||||
|
||||
## The trap
|
||||
chars/4 under-counts the rendered chat template by ~3 % on code/agent traffic. If contextWindow == engine ceiling, requests with est > 4096/0.03 ≈ 137k are guaranteed 400s below the compaction threshold (65-85 % of a 200k window). maxTokens ≤ 16,384 hides this; 45,056 exposes it.
|
||||
|
||||
## Fix procedure
|
||||
1. Compute required slack: maxTokens + 0.10 × contextWindow (robust to est error).
|
||||
2. Raise the engine ceiling to ≥ contextWindow + slack (native window is the natural choice: Qwen3.8-Flash-Next 262,144). Halogen: pool must stay slots × CTX (`HALOGEN_KV_POOL_POSITIONS`, ~28 KiB/position); check memory ledger.
|
||||
3. Keep harness contextWindow below the ceiling by ≥ slack (217,088 = 212×1024 for a 262,144 ceiling with maxTokens 45,056).
|
||||
4. Verify compaction itself fits its target (remoteEndpoint slot ≥ thresholdPercent × contextWindow − keepRecentTokens(20k) + overhead).
|
||||
5. Update env + seam test + models.yml + AGENTS.md together; a single-file change re-creates the drift.
|
||||
|
||||
## Verification
|
||||
Probe a prompt at ≥ 180k tokens with max_tokens = the clamped value; expect completion, not 400.
|
||||
@@ -0,0 +1,62 @@
|
||||
---
|
||||
name: omp-model-id-rename-cutover
|
||||
description: Rename an OMP fleet model id in models.yml and prove every reference and endpoint still resolves before commit/push.
|
||||
---
|
||||
|
||||
# OMP model id rename cutover
|
||||
|
||||
Renaming a model id in `~/.omp/agent/models.yml` touches three layers: the registry, every
|
||||
reference in `agent/config.yml`, and the wire contract with the serving endpoint.
|
||||
|
||||
## 1. Find every reference
|
||||
Provider names are labels; ids are referenced from `config.yml` (`modelRoles.*`,
|
||||
`task.agentModelOverrides.*`, `compaction.modes`), from other models' `compactionModel`,
|
||||
and from docs/memory. Grep the tracked set:
|
||||
|
||||
```bash
|
||||
git ls-files -z | xargs -0 grep -ln 'OLD_ID'
|
||||
```
|
||||
|
||||
`agent/memories/**` and `omp-docs/log.md` are historical records: leave them alone.
|
||||
`agent/models.yml` comments that name the *service* (e.g. `frank-large.env`) stay — the
|
||||
systemd unit names do not change with the harness id.
|
||||
|
||||
## 2. Prove reference integrity (throwaway, do not commit)
|
||||
Parse both YAMLs and cross-check. The provider label contains a space and parentheses
|
||||
(`llamacpp (dankserver)`), so use an explicit pattern; a `split(/\s+/)` yields false
|
||||
unresolved hits like `(dankclaw)/frank-dragoon`.
|
||||
|
||||
```python
|
||||
pat = re.compile(r"llamacpp \([^)]+\)/[A-Za-z0-9._-]+")
|
||||
known = {f"{n}/{m['id']}" for n, pv in models["providers"].items() for m in pv["models"]}
|
||||
# walk cfg.modelRoles / cfg.task / compactionModel, flag tokens not in known
|
||||
```
|
||||
|
||||
Expect: zero unresolved refs, and every `compactionModel` present in `known`.
|
||||
|
||||
## 3. Prove the endpoint accepts the new id
|
||||
`/v1/models` reports the SERVED name (radiance echoes the request id, halogen and
|
||||
llama.cpp echo their own). Then send a 4-token completion with the NEW id against each
|
||||
provider `baseUrl`:
|
||||
|
||||
```bash
|
||||
curl -s http://HOST:PORT/v1/chat/completions -H 'content-type: application/json' \
|
||||
-d '{"model":"NEW_ID","messages":[{"role":"user","content":"say OK"}],"max_tokens":4}'
|
||||
```
|
||||
|
||||
200 + `reasoning_content` = the rename is harness-side only. All three engines on the
|
||||
Dank fleet ignore the request `model` field.
|
||||
|
||||
## 4. Fit check against the served window
|
||||
Read the served window (`/v1/models` → `context_length` / `meta.n_ctx`) and confirm the
|
||||
declared `contextWindow + maxTokens` fits the slot/pool the engine admits against.
|
||||
Compaction fires at `compaction.thresholdPercent` of `contextWindow`; the clamp is
|
||||
`max_tokens = min(maxTokens, contextWindow − est − 4096)` and the chars/4 estimate
|
||||
under-runs ~3 % on code.
|
||||
|
||||
## 5. Commit shape
|
||||
One commit for `models.yml` (ids + windows + sampling), one for `config.yml` (role map,
|
||||
compaction endpoint). Note in the message which wire names stay unchanged, and record
|
||||
the rename in `omp-docs/log.md` (newest section first). Run `scripts/kb-lint.sh` before
|
||||
committing docs. The `.githooks/pre-commit` secret gate runs automatically; never
|
||||
`--no-verify`.
|
||||
@@ -0,0 +1,67 @@
|
||||
---
|
||||
name: omp-system-prompt-retrieval
|
||||
description: "Retrieve the exact system prompt an omp session is running: /dump while live, RPC get_state re-render, and the shipped template/override source paths."
|
||||
---
|
||||
|
||||
# Retrieve omp's live system prompt
|
||||
|
||||
Three routes, in fidelity order. Verified on omp 18.4.10 (linux/x64).
|
||||
|
||||
## 1. `/dump` — the only exact capture of a LIVE session
|
||||
TUI slash command: "Copy session transcript to clipboard (and write LLM request JSON to tmp)".
|
||||
The request JSON is the wire-true payload (system blocks + tool schemas). Path is printed in
|
||||
the `/dump` output; then:
|
||||
|
||||
```bash
|
||||
jq -r '.system[] | strings // .system' <file> # anthropic shape
|
||||
jq -r '.messages[] | select(.role=="system") | .content' <file> # openai shape
|
||||
```
|
||||
|
||||
**The prompt is never persisted in the session JSONL.** Session entries are only
|
||||
`title, session, model_change, thinking_level_change, message, custom(tool_execution_start),
|
||||
model_usage`. So a past session's prompt is unrecoverable from disk — capture live, or re-render.
|
||||
|
||||
## 2. RPC `get_state` — re-render without a model call
|
||||
`omp --mode rpc` answers `{"type":"get_state"}` with `data.systemPrompt: [blocks]` and
|
||||
`data.dumpTools` (spec: `omp://rpc.md` "get_state payload"). Drive it over stdio:
|
||||
|
||||
```python
|
||||
cmds = json.dumps({"id":"p1","type":"negotiate_protocol","protocolVersion":2}) + "\n" \
|
||||
+ json.dumps({"id":"s1","type":"get_state"}) + "\n"
|
||||
p = subprocess.run(["omp","--mode","rpc","--no-session","--no-pty"],
|
||||
cwd=TARGET_CWD, input=cmds, capture_output=True, text=True, timeout=280)
|
||||
```
|
||||
|
||||
Closing stdin shuts it down cleanly (exit 0), no model call, no session file written.
|
||||
|
||||
- Parse stdout line-by-line. Protocol v2 wraps logical frames >1 MiB in `rpc_chunk` frames:
|
||||
key on `chunkId`, require all `count` indices, `base64.b64decode` and concatenate in `index`
|
||||
order, then `json.loads` — otherwise a large prompt is silently lost.
|
||||
- Shape measured on a dotfiles repo cwd: 2 blocks, ~45 KB / ~391 lines —
|
||||
block 0 = static instruction block (RFC 2119 line, `§ Role`, Engineering/Tone/Escalation,
|
||||
tool inventory, `xd://` catalog), block 1 = footer opening `<project-context>` +
|
||||
`<critical>` + the cwd's `AGENTS.md`.
|
||||
- `--no-extensions` changes nothing in the blocks; tool count rides in `dumpTools`, not the prompt.
|
||||
- `cwd` matters: the footer is cwd-derived (workstation block, context files, skills, rules).
|
||||
|
||||
## 3. Sources (edit here, not in a dump)
|
||||
| Layer | Path |
|
||||
|---|---|
|
||||
| Default instruction template (Handlebars, block 0) | `<pkg>/src/prompts/system/system-prompt.md` |
|
||||
| Footer | `<pkg>/src/prompts/system/project-prompt.md` |
|
||||
| Plain-custom template | `<pkg>/src/prompts/system/custom-system-prompt.md` |
|
||||
| Repo rules (into footer) | the cwd's `AGENTS.md` |
|
||||
| Overrides | `--system-prompt`, `--system-prompt-template <path>`, `--append-system-prompt`; discovered `SYSTEM.md`, `SYSTEM_TEMPLATE.md`, `APPEND_SYSTEM.md`, `PERSONALITY.md`, `TITLE_SYSTEM.md` |
|
||||
|
||||
`<pkg>` = `~/.bun/install/global/node_modules/@oh-my-pi/pi-coding-agent` (resolve via
|
||||
`readlink -f $(command -v omp)`). Docs: `omp://system-prompt-customization.md`.
|
||||
|
||||
Precedence: explicit CLI flag > project `SYSTEM.md` > project `SYSTEM_TEMPLATE.md` >
|
||||
user `SYSTEM.md` > user `SYSTEM_TEMPLATE.md`. A same-scope literal always beats a discovered
|
||||
template, so a `SYSTEM_TEMPLATE.md` is inert while a `SYSTEM.md` sits beside it.
|
||||
|
||||
## Trap
|
||||
NEVER paste a rendered `/dump` into `SYSTEM_TEMPLATE.md`: a render freezes settings, tool
|
||||
catalogs, and mounted-device data. Copy the Handlebars **source** and diff against the shipped
|
||||
path after every omp upgrade. Plain files (`SYSTEM.md`, `APPEND_SYSTEM.md`) are inserted
|
||||
verbatim — `{{cwd}}` and friends stay literal there; only the template route compiles Handlebars.
|
||||
@@ -0,0 +1,29 @@
|
||||
---
|
||||
name: openwebui-picker-acl-repair
|
||||
description: "Fix OpenWebUI 0.6 \"No results found\" / Model not found in the model picker for non-admin users (Dankclaw fleet) after a connection rebuild."
|
||||
---
|
||||
|
||||
# OpenWebUI 0.6 picker ACL repair (Dankclaw)
|
||||
|
||||
Symptom: admin sees models in OpenWebUI; regular users get "No results found" in the picker, or chat answers "Model not found". Cause: an OpenAI-connection rebuild (config keys `openai.api_base_urls` / `openai.api_configs` in the `config` table) that dropped the ACL pair for the model.
|
||||
|
||||
## Law
|
||||
A non-admin user sees an external OpenAI-connection model only if BOTH exist in `webui.db`:
|
||||
1. `model` table (singular) row: `id` == upstream model id, `base_model_id` NULL, `is_active` 1, `meta` JSON (description etc.).
|
||||
2. `access_grant` row: `(resource_type='model', resource_id=<model id>, principal_type='user', principal_id='*', permission='read')`.
|
||||
|
||||
The aggregation (`utils/models.py get_filtered_models`) drops info-less models for `role=user`; admins bypass. `get_all_models` responses are served from `model_ids` in `api_config` when non-empty (no upstream fetch). An alias row (`id` != upstream, `base_model_id` set) renders in the picker but chat fails "Model not found" — id must equal the upstream id.
|
||||
|
||||
## Steps
|
||||
1. Reproduce as the affected role: in a browser tab, `fetch("/api/models", {headers:{Authorization:"Bearer "+localStorage.token}})` → `data: []` while an admin API key (`api_key` table, column `key`, `sk-…`) lists the model.
|
||||
2. DB at `/home/dank/Containers/open-webui/data/webui.db` (host; no sqlite3 CLI on Dankclaw — use host `python3` heredoc over ssh; container path `/app/backend/data`).
|
||||
3. Insert/repair the pair (uuid4 id, epoch-ms `created_at`):
|
||||
- `model`: id/name = upstream id, base_model_id NULL, is_active 1, user_id = admin uid (role='admin' in `user` table), meta `{"description": "…"}`.
|
||||
- `access_grant`: shape above.
|
||||
4. Cache: `MODELS_CACHE_TTL` defaults to 1 s — no restart needed for lists; a restart (`systemctl --user restart open-webui`, rootless) only if config edits were cached oddly. An OpenWebUI restart invalidates browser JWTs (401) — expect a silent SSO re-round-trip; kill stale tabs first so OAuth `state` cookies don't race (MismatchingStateError → clear cookies/storage, one clean flow).
|
||||
5. Verify end-to-end: pick the model in the UI, send a short exact-match prompt, confirm the reply; check `/api/models` as the user role lists the id.
|
||||
|
||||
## Related knobs
|
||||
- User API keys (User menu → Settings → Account → API keys → Show → Create new secret key) require `user.permissions` JSON `features.api_keys=true` in `config` (default false).
|
||||
- New SSO signups land as role `pending` (`ui.default_user_role=pending`) — activate: `UPDATE user SET role='user' WHERE email LIKE …`. Table is `user`, singular; no `active` column.
|
||||
- Anubis gates `ai.franksdank.cloud` for headless browsers (webdriver detection): drive direct `:3005` for UI work, or emulate a desktop Chrome UA.
|
||||
@@ -1,29 +0,0 @@
|
||||
---
|
||||
name: quadlet-v1-generator-quirks
|
||||
description: "When writing or deploying Quadlet .container units on podman 5.x (v1 generator, e.g. Dankclaw), avoid keys the v1 generator hard-fails on and verify generated ExecStart before starting; generated units cannot be enabled."
|
||||
---
|
||||
|
||||
# Quadlet v1 generator quirks (podman 5.x)
|
||||
|
||||
The podman 5.x user generator (`/usr/lib/systemd/user-generators/podman-user-generator`, quadlet **v1**) behaves differently from the v4 docs. Verified live on Dankclaw 2026-08-28 while deploying the authentik trio.
|
||||
|
||||
## Hard-fail keys (unit silently fails to generate — `systemctl --user cat <unit>` shows "could not be found")
|
||||
- Any dash-prefixed option line in `[Container]`: `--shm-size=512m` etc. → generator logs `unsupported key '--shm-size' in group 'Container'` and skips the whole unit. Omit them; accept podman defaults.
|
||||
- `Command=` is unsupported — use `Exec=` (that is what this generator accepts).
|
||||
|
||||
## HealthCmd space-splitting
|
||||
The v1 generator splits `HealthCmd` on spaces into the healthcheck argv array (`podman inspect --format {{.Config.Healthcheck.Test}} <ctr>` shows `[CMD-SHELL pg_isready -d ... ]`). A quoted probe like `/bin/sh -c "pg_isready ..."` lands as argv token `"pg_isready` → command-not-found → container reads **unhealthy while the DB is actually up**. Rule: HealthCmd must contain only space-free argv tokens, e.g. `HealthCmd=pg_isready -d authentik -U authentik` or `python /manage.py check --database default`.
|
||||
|
||||
## Generated units cannot be enabled
|
||||
`systemctl --user enable <generated-unit>` fails with "unit is transient or generated". Boot persistence comes from `[Install] WantedBy=default.target` inside the `.container` file, honored by the generator at user-manager start; linger must stay on for the deploy user. Deploy handlers: install unit → `systemctl --user daemon-reload` → `systemctl --user start`; no `enable`.
|
||||
|
||||
## Probe a new key before shipping it
|
||||
Drop a throwaway `probe.container` into `~/.config/containers/systemd/`, `systemctl --user daemon-reload`, then check the generator verdict:
|
||||
|
||||
```bash
|
||||
journalctl --user --since "1 min ago" -o cat | grep -i "unsupported key"
|
||||
systemctl --user cat probe.service # shows the generated ExecStart verbatim
|
||||
rm ~/.config/containers/systemd/probe.container && systemctl --user daemon-reload
|
||||
```
|
||||
|
||||
If a healthcheck ever reads unhealthy but logs look fine, suspect this quirk first: compare the baked `Healthcheck.Test` array against the intended command before believing the status.
|
||||
@@ -0,0 +1,32 @@
|
||||
---
|
||||
name: radiance-frank-primary-benching
|
||||
description: "Bench frank-primary (radiance) decode/prefill via the scratch-container harness: method, noise floor, and the MNBT/HOST_POOL/envelope laws to respect before proposing any flag change."
|
||||
---
|
||||
|
||||
# Benching frank-primary (radiance)
|
||||
|
||||
## Method (works as of 2026-10-06)
|
||||
- Entry point: `bash autoresearch.sh` (dev box) → scp's `scripts/frank-primary.env` + `scripts/autoresearch-frank-primary.sh` to dankserver → the bench script ON dankserver: stops frank-primary (relinked unit: `stop`, never `mask`), boots scratch container `arbench` on the candidate env, polls /health (binds after full load, ~110 s cold), runs ONE greedy completion (fixed ~140k-char repo-corpus prompt, seed 1234 → 43,627 prompt tokens, 256 decode tokens, temp 0), parses the engine `timings` object, tears down, and restores prod via a trap.
|
||||
- Metrics from `timings`: `prompt_per_second` (cold prefill), `predicted_per_second` (decode at depth = (predicted_n-1)/predicted_ms), `prompt_ms` (TTFT), `draft_n`/`draft_n_accepted` (acceptance).
|
||||
- Fresh prefix-cache tier per run (`/data/kvcache/arbench`) so cold prefill is cold; prod's tier untouched.
|
||||
- Bench locally on dankserver (127.0.0.1) — no LAN jitter. Run-to-run ~4-5 min.
|
||||
|
||||
## Noise floor (measurement law)
|
||||
- Decode: ~±3% on IDENTICAL config (33.06 / 32.03 / 30.90 observed). Single-run deltas <3% are not evidence; re-run before keeping.
|
||||
- Prefill: ~±2% (453/448/457 identical config).
|
||||
|
||||
## Laws at the seqs=3 / maxlen=200k / fp8-KV / MTP-on shape (2×R9700, TP=2)
|
||||
- **MNBT is a sharp resonance at 2048**: 2560 → decode 2.6-7.7 t/s (acceptance stays normal, verify stalls ~130 ms/step, generation truncates); 1536 → decode 24.7 + 22% prefill loss + truncation; 4096 → boot failure at seqs=3 (boots at seqs=2, -2.6% only). DO NOT move MNBT off 2048.
|
||||
- **HOST_POOL_MIB 40960 is the load-bearing floor**: 36864 fails boot; 49152 boots but decode-flat.
|
||||
- **EXPERT_VS_CACHE 0.82 is bracket-confirmed**: 0.78 fails boot; 0.84 costs 8.9% (toward the documented 0.87 collapse cliff).
|
||||
- **TP_WIRE wht6 is mandatory** at seqs=3: bf16 (lossless) fails boot.
|
||||
- **SPEC_TOKENS**: depths 1-3 statistically tied on decode; 0 = -6.8% (real), 4 = -22.5% (acceptance collapses to 0.50). Depth 2 nominal peak.
|
||||
- MAX_MODEL_LEN 262144 (native) costs 2.5% — the larger KV pool steals resident-weights VRAM at expert_tiered.
|
||||
- GPU_HEADROOM 48 vs 96: no effect.
|
||||
|
||||
## Hazard: abandoned bench contamination
|
||||
- If a prior bench run is skipped/abandoned mid-flight, the next scratch boot may race the leftover engine and produce garbage numbers. Before trusting any collapse measurement: `podman ps -a` on dankserver — confirm the previous arbench is gone and frank-primary is active — then reproduce the collapse on a clean boot before writing it into the record.
|
||||
|
||||
## State
|
||||
- HEAD on the autoresearch branch: spec=2, seqs=3, maxlen=200k (df6b5bb6e807). All spec 1-3 × seqs 2-3 configs decode within 30.9-33.1 — statistically tied. Frank's decision pending: keep spec=2/seqs=3 or revert to his spec=3/seqs=2 (83% acceptance, likely better for real multi-turn traffic).
|
||||
- Production verified healthy (health 200) after every run via the harness trap.
|
||||
@@ -0,0 +1,31 @@
|
||||
---
|
||||
name: radiance-precision-selection
|
||||
description: "Choose tensor precisions for a radiance (StillDeadcode) rad-convert recipe on RDNA4: what is native, what is supported, what the hardware rewards."
|
||||
---
|
||||
|
||||
# Radiance precision selection
|
||||
|
||||
Use when writing/choosing a `rad-convert` recipe for radiance on RDNA4 (gfx1201), or when arguing why a model was converted a particular way.
|
||||
|
||||
## Project map (which repo is authoritative)
|
||||
- `codeberg.org/StillDeadcode/radiance` — standalone engine (C++/HIP). Recipes: `data/recipes/*.recipe`; docs: `docs/{GUIDE,MOE-W4,KERNELS,PLUGIN,TOOLS,OPS,ARCHITECTURES,DOCKER}.md`. Prefer the RAW form of Codeberg URLs when reading: `/raw/branch/main/<path>`.
|
||||
- `codeberg.org/StillDeadcode/libr4d` — kernel library; entry points encode exact geometry and reject mismatch (`attn_decode_h256_gqa6_fp8kv`, `gdn_chunk_scan_k128_v128_c64_bf16`, `ar_oneshot_2rank_exact`). Kernel names are the ground truth for which dtypes have a fast path.
|
||||
- `codeberg.org/StillDeadcode/vllm-radiance` — deprecated vLLM wrapper; README still carries the best prose on the FP8 dispatcher and kernel list.
|
||||
|
||||
## Precision ladder
|
||||
| Precision | Role |
|
||||
|---|---|
|
||||
| block-FP8 (E4M3 + bf16 block scales) | native tuned GEMM/attention/KV — "preshuffled block-FP8 dispatcher" |
|
||||
| MXFP4 (OCP group-32) / W4A8 | supported weight packing; activations dynamically quantized to FP8 E4M3, compute is FP8 WMMA |
|
||||
| bf16 | KV (`auto`), skinny GEMMs, Gated-DeltaNet kernels |
|
||||
| int8 / int4 | draft head and integer GEMM paths |
|
||||
|
||||
## Decision rules
|
||||
- Prefer block-FP8 for anything the engine qualifies: generic upstream AITER linear in place of the FP8 dispatcher measured **9–11 % slower**; all upstream-qualified models are FP8 (Qwen3.8-27B-FP8, Qwen3.6-27B-FP8, Qwen3.6-35B-A3B-FP8, Gemma-4-31B-it-FP8).
|
||||
- Weight packing is a *storage* choice, activation dtype is the *compute* choice: MXFP4 buys bytes, not an FP4 compute path.
|
||||
- The matrix core is integer-biased: f16 WMMA is half the int8 rate and a quarter of the int4 rate ⇒ int4 > int8 > f16. FP8/INT8 kernels must fuse double-K WMMA to saturate the 128-bit memory interface (GPUOpen WMMA guide part 2) — a "slow" quantized kernel usually means a missing K-fused variant, not an inherent dtype cost.
|
||||
- Keep low-bit tricks on the draft head (int4/int8) where verification tolerates them; keep lm_head and hard-selection tensors (QSA indexer top-k) at high precision — the shipped recipes quantify the cost (int8 lm_head beats E4M3 by ~0.6 top-1 points).
|
||||
|
||||
## Verification
|
||||
- Read the recipe's own header comments: radiance recipes state measured KL / top-1 / prefill-decode deltas for their choice, and name the `--reuse` inputs.
|
||||
- Confirm the kernel exists before promising a dtype: grep libr4d entry points for the dtype suffix (`fp8kv`, `bf16`, `exact`).
|
||||
@@ -0,0 +1,26 @@
|
||||
---
|
||||
name: radiance-quantize-and-score
|
||||
description: "Design a rad-convert recipe for a radiance (StillDeadcode) MoE checkpoint, calibrate it, and score it with the KL mode — the procedure plus the servable-format law."
|
||||
---
|
||||
|
||||
# Radiance: quantize, calibrate, score
|
||||
|
||||
For a HuggingFace checkpoint → `.rad` on the radiance engine (Codeberg StillDeadcode/radiance).
|
||||
|
||||
## 1. Know the servable formats before writing a rule
|
||||
The block/plugin serves a fixed list; the converter refuses an encoding no kernel reads (`docs/TOOLS.md:184`). For qwen4exp MoE experts: `fp8a8, w4a8, w4a8h, bf16, w4nla8h, w4nl64a8h, w4nl32a8h, w5nl64a8h` (`arch/common/rad_block_moe_fp8.h` `kServed`); dense: `fp8a8, i8a8, fp8a16, mxfp4a8, bf16`, plus the `u2` draft head. `fp6_*`/`u6` encode but never serve. Read `kServed` in source — never trust a code-name list from docs prose.
|
||||
|
||||
## 2. Recipe file
|
||||
Header `#` lines are stored in the container; first matching rule wins, so put `cast dtype=bf16` protected-expert rules and projection-specific rules **ahead of** the general `blk.*.ffn_*_exps.*.weight` rule. Unmatched weights are served as the checkpoint holds them (bf16). Verify names against radiance's internal stems (`ffn_gate_up_exps`, `ffn_down_exps`, `attn_qg`, `ssm_inz`, `ple_ngram`, `output.weight`), not the HF names.
|
||||
|
||||
## 3. Calibrate (GPTQ/awq/autoround/paroquant need `$CALIB`)
|
||||
Bootstrap: `rad-convert` with plain `rtn` → serve with `RADIANCE_CALIB_DIR` (+`RADIANCE_CALIB_TOKENS`) and drive prefill → re-convert with `--recipe … --reuse <bootstrap.rad>` (`--reuse` copies every weight whose encoding did not change). A BF16 checkpoint may be calibrated as it ships: convert with no recipe + `--max-ctx 4096`, serve `--placement expert_tiered --weights-disk-tier`, MNBT large (a step costs the same at 1 token as at 32K). Corpus must look like production traffic; score on a held-out slice.
|
||||
|
||||
## 4. Score
|
||||
`radiance --model BF16.rad … --kld-record DIR --kld-corpus C.jsonl` (exact wire, bf16 KV) once, then `--kld-ref DIR --kld-out R.json` per candidate. Report = mean/median/p99/p99.9/p99.99 KL, top-1 agreement, both perplexities, **per `source`** — label the corpus's `source` field to get per-turn-type scopes. KLD corpus disjoint from calibration. Keep `--max-num-batched-tokens ≤ 8192` (the logits buffer is ~2 GiB/rank per 4096 tokens).
|
||||
|
||||
## 5. Set the gate from the published ladder, not from a wish
|
||||
Compare against `docs/MOE-W4.md`'s table for the arch (served tier / heavier tier / bf16 floor). A threshold below the bf16 floor rejects every quantized build by construction. Escalation ladder reuses the same Grams (Hessians are grid-independent) → escalate by re-converting only.
|
||||
|
||||
## 6. Verify before serving
|
||||
`rad-tune` at the served shape (TP, spec depth) with a persistent cache mount; read the boot ledger for residency; run the fleet probes.
|
||||
@@ -0,0 +1,18 @@
|
||||
---
|
||||
name: radiance-w4-campaign
|
||||
description: "Run a radiance GPTQ-w4 quantization campaign on dankserver: calibration → protected experts → convert → KLD gate → deploy, with the engine-pin and MNBT laws."
|
||||
---
|
||||
|
||||
Pipeline for quantizing a BF16 Qwen3.8-Flash-Next checkpoint into a servable `qwen4exp` `.rad` on dankserver (frank-primary). Reference run: `serverAppConfigs-docs/LLM/frank-primary-radiance-swift15-w4dn8.md`; scripts in `scripts/` (`run-swift15-bf16.sh`, `calib-push.py`, `derive-protected-experts.py`, `apply-protected-cast.sh`, `convert-swift15-w4dn8.sh`, `kld-gate-swift15.sh`, `fleet-probes-swift15.sh`).
|
||||
|
||||
Sequence:
|
||||
1. **Fetch** checkpoint with per-shard LFS sha256 verification (`-f`, byte counts, no token on argv).
|
||||
2. **BF16 container**: `rad-convert` no-recipe (pure copy) — serves as the KLD reference engine.
|
||||
3. **KLD reference**: record logits on the SERVE pin (1.0.13) at MNBT 32768, `--max-model-len 16384`, corpus ≤16k prompts. Writes atomically at completion (`rows:` line = receipt). Never scp a running bash script (lazy-read law).
|
||||
4. **Calibration**: engine **1.2.0** (only pin with `RADIANCE_CALIB_TOKENS`); push ~10M tokens of the model's own distribution; per-expert down-proj Grams land as RADGRAM2 files (`gram.r<rank>.blk.L.ffn_down_exps.<e>.bin`). Grams are engine-invariant across pins — the version boundary is by design.
|
||||
5. **Protected experts**: derive from gram energy (share ≥3 % AND ≥3× routing share), splice between `# PROTECTED-BEGIN/END` markers of the recipe. Plugin law: ONE encoding per expert table; only whole bf16 experts legal beside it — no gate_up/down split.
|
||||
6. **Final convert** on the SERVE pin (CPU GPTQ, hours; disk gate ≥146 GiB free first).
|
||||
7. **KLD gate** on the serve pin: **MNBT ≤ 2048** (w4 streams in place only there; 32768 → `DID NOT FIT ... stored form is not the container's`), pool 40960, fp8 KV matching the production KV dtype. Gate: mean KL ≤ 0.03, top-1 ≥ 95 %.
|
||||
8. **Deploy** `./scripts/deploy.sh dankserver --service frank-primary`; verify `/v1/models` id = served-model-name (alias independent of file — harness models.yml untouched), boot ledger, streamed ≥128-token decode probe, fleet probes (vision needs ≥256-token budget + `reasoning_content` fallback).
|
||||
|
||||
Laws: stop frank-primary for every GPU phase (relinked unit: stop, mask fails); verify docs against `podman inspect <name> --format {{join .Config.Cmd " "}}` (rootless, as dank) before writing shapes; ff-pull fails on untracked files the merge tracks — trash first.
|
||||
@@ -0,0 +1,29 @@
|
||||
---
|
||||
name: radiance-w4-convert-playbook
|
||||
description: "Convert a BF16 Qwen3.8-Flash-Next checkpoint to a GPTQ-w4 radiance .rad: recipe law, calibration tap, derive-protected-experts, convert gates, KLD gate shape and flags."
|
||||
---
|
||||
|
||||
# Radiance w4 convert playbook (dankserver)
|
||||
|
||||
End-to-end recipe for rad-convert GPTQ w4 on the qwen4exp arch (engine pin `stilldeadcode/radiance@sha256:4c3f48...` = 1.0.13 for convert/KLD/serve; 1.2.0 `sha256:665775...` ONLY for the calibration tap).
|
||||
|
||||
## Pipeline
|
||||
1. **Calibration corpus** ~10M tokens (`make-calib-corpus.py`, seed-pinned; keep `kld.jsonl` for the gate — regenerate deterministically if lost).
|
||||
2. **KLD reference on the BF16 container** (no-recipe convert, `--kld-record`), same engine as scoring.
|
||||
3. **Calib serve on 1.2.0**: `RADIANCE_CALIB_DIR` + `RADIANCE_CALIB_TOKENS=10000000`; tap in `qwen4exp_fp8.so` writes `gram.r<rank>.blk.L.ffn_down_exps.<e>.bin` (RADGRAM2, 32 B header + k×k f32); DONE mark `calibration wrote`.
|
||||
4. **Derive protected experts** from the Grams: energy share ≥3 % AND ≥3× routing share, even count per rank half.
|
||||
5. **Recipe law**: ONE encoding per expert table — all routed experts through GPTQ w4 (`libquant: GPTQ quantises 'blk.N.ffn_gate_up_exps' against M calibration file(s)`); PROTECTED section = whole bf16 experts ONLY, spliced between `# PROTECTED-BEGIN/END` markers. The $CALIB env var is expanded by rad-convert.
|
||||
6. **Convert** host-backend (no GPU; prod-compatible), ~6 min/layer × 49, CPU-saturating. In-container paths must include subdirs (mount prefix strip). Gate: rad-info shows gptq rule; size floor from geometry, not by analogy.
|
||||
|
||||
## KLD gate shape (1.0.13)
|
||||
`--placement expert_tiered --weights-disk-tier --host-pool-mib 40960 --deterministic --gpu-headroom-mib 512 --num-speculative-tokens 0 --kv-cache-dtype fp8 --max-model-len 16384 --ngram-placement ram --kld-ref DIR --kld-out FILE --vram-kv-mib 2048 --max-num-seqs 4 --max-num-batched-tokens 3072`.
|
||||
- `--kld-ref` takes NO `--kld-corpus` (rejects the pair).
|
||||
- w4 arena ≈ 262.5 MiB per 256 tokens of MNBT — 32768 needs 33 GiB of VRAM.
|
||||
- w4 gate_up experts can't disk-tier stream (stored ≠ served form) — pool must hold ~12 GiB/rank.
|
||||
- Gate: mean KL ≤ 0.03 AND top-1 ≥ 95 % vs reference.
|
||||
|
||||
## Pitfalls
|
||||
- Editing a running bash script corrupts its execution (lazy byte-offset reads) — sync after the run.
|
||||
- `pkill -f <pattern>` kills your own ssh watcher if the pattern matches the ssh command line.
|
||||
- zsh on dankserver eats nested quoting — write scripts locally, scp, `sh /tmp/x.sh`.
|
||||
- Radiance image has no bash/strings: `--entrypoint sh`, `grep -ac` with positive control.
|
||||
@@ -0,0 +1,22 @@
|
||||
---
|
||||
name: static-smoke-server-verify
|
||||
description: Verify static frontend edits on a scratch http.server + headless browser without stale-server/browser-cache false negatives
|
||||
---
|
||||
|
||||
## When
|
||||
Serving a copied frontend tree (e.g. `shared/homepage/` → `/tmp/hp-smokeN/`) via `python3 -m http.server` and verifying with the managed headless browser.
|
||||
|
||||
## Procedure
|
||||
1. Kill any prior scratch server FIRST: `pkill -f "http.server <port>"; sleep 0.5`. A second launch on an occupied port exits silently and the OLD server keeps serving the OLD copy.
|
||||
2. Fresh copy each wave: `rm -rf /tmp/hp-smokeN && cp -r <tree> /tmp/hp-smokeN`; for single-file resyncs `cp` the changed files into the served dir.
|
||||
3. Start server from the served dir; prove the SERVED bytes are new before browser work:
|
||||
`curl -s http://127.0.0.1:<port>/<page> | grep -c <marker-only-in-new-file>`
|
||||
If curl is green but the browser shows old content, it is browser disk cache (http.server sends no Cache-Control).
|
||||
4. Cache-bust every browser navigation: `?v=<timestamp>` query on page URLs.
|
||||
5. Assert behaviorally in `tab.evaluate` (section id order, computed styles, `naturalWidth` of imgs, class presence), screenshot for the visual record, then `tab.close`.
|
||||
|
||||
## Cleanup
|
||||
`pkill -f "http.server <port>"`; move smoke trees to `~/.AGENTTRASH/YYYYMMDD-topic` (never `rm`).
|
||||
|
||||
## Harness quirk
|
||||
bash async calls with `name`+`timeout` may be rejected ("Service mode does not accept async or timeout"); plain `async` works, `ready: {timeout}` for services.
|
||||
@@ -0,0 +1,27 @@
|
||||
---
|
||||
name: subagent-yield-failure-recovery
|
||||
description: "Recover a subagent's full structured report when it exited without yielding (SYSTEM WARNING, structured output unavailable)"
|
||||
---
|
||||
|
||||
A subagent that solved the task but exited without calling `yield` (status `failed`, "exited without calling yield tool") usually still has its full payload in the transcript. Recover it; do not re-run the subagent.
|
||||
|
||||
## Procedure
|
||||
1. `read agent://<id>` — the preview truncates long lines (~768 chars); the full JSON draft appears in the transcript, not here.
|
||||
2. Dump the raw transcript with line truncation off and persist it:
|
||||
```js
|
||||
const raw = await read("history://<id>:raw");
|
||||
await write("/tmp/hist.txt", raw);
|
||||
```
|
||||
3. Extract the report string. The transcript embeds the agent's draft as a JS object literal (`report: "...."` with escaped quotes/newlines), repeated across retries — take the longest match, then JSON-unescape:
|
||||
```python
|
||||
import re, json
|
||||
raw = open('/tmp/hist.txt', encoding='utf-8', errors='replace').read()
|
||||
best = max((m.group(1) for m in re.finditer(r'report: "((?:[^"\\]|\\.)*)"', raw)), key=len)
|
||||
open('report.md','w').write(json.loads('"' + best + '"'))
|
||||
```
|
||||
Same pattern works for `summary:`/`files:` fields.
|
||||
4. Treat recovered content as UNVERIFIED scout output: cross-check high-stakes claims at primary sources before publishing (scouts can conflate similarly-named projects; one recovery revealed "arc53's quantization pruner" when the tool was actually `p-e-w/heretic`, an abliteration tool).
|
||||
|
||||
## Why
|
||||
- `agent://<id>` and the task-result preview truncate per line; only `history://<id>:raw` carries the full text.
|
||||
- Re-spawning the subagent wastes minutes and the cap slot; the work product already exists in the transcript.
|
||||
@@ -0,0 +1,28 @@
|
||||
# Environment & Tooling Memory (Frank workstation / blankpad)
|
||||
|
||||
## Persistent facts
|
||||
- User: `blank`. Home: `/home/blank`. Dotfiles repo: `~/.omp` (Gitea `dank/command-center` at 192.168.0.143:3000, local `git push` pattern). `serverAppConfig` cloned to `~/Projects/serverAppConfig`.
|
||||
- Desktop: Noctalia (Niri-based DE at `/usr/bin/noctalia`). Configs `~/.config/noctalia/config.toml` + `noctalia.kdl`; IPC via `noctalia msg <cmd>` (panel-toggle, session, lock, launcher). App templates in `/usr/share/noctalia/assets/templates/`. Palette generated via `noctalia theme [image]`. GTK accent #b267e4.
|
||||
- Ghostty: `theme = noctalia`; text AA via `freetype-load-flags` (not `font-hinting`). Alacritty: `font.antialias=false`.
|
||||
- Niri display config: `~/.config/niri/cfg/display.kdl`. Transform values are absolute; in portrait, a 180° flip = swap 90° ↔ 270°.
|
||||
- Hardware: Lenovo laptop eDP-1 (1920x1080, scale 1.25) + two PNP(XYM) PM1601 externals (1920x1200, scale 1, portrait). Current blankpad layout: DP-2@x=0 → HDMI-A-1@x=1200 → eDP-1@x=2400.
|
||||
- Frank desktop: cursor size 36, capitaine-cursors theme, Authenik 2026.8. Authenik caches claims per session (1-24h TTL) → silent 403s after group add.
|
||||
- Inference: `handy` app (`/usr/bin/handy`, hub op=start, application=/usr/bin/handy, pty=true, persist=true) with 69 catalog models. FLM NPU gemma4-it on port 7771 — PNG/JPEG only (no WebP), needs `thinkingLevelMap` in models.json, `.xclbin` symlinks under `~/.config/flm/xclbins/`.
|
||||
- Halogen runtime (frank-large): baked-in chat template, no file swap; accepts `chat_template_kwargs` (enable_thinking, preserve_thinking, reasoning_effort). Sampler defaults: temp=1.0, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0. `preserve_thinking=true` in `agent/models.yml`.
|
||||
- `sudo usermod -aG input blank` was required for `/dev/input` device access (660 root:input).
|
||||
- Compositor AA in niri 26.04 is limited to corner rounding via `niri_rounding_alpha()`; blur/noise settings cause softness, not AA. Client-side text AA is not controlled by niri.
|
||||
|
||||
## Durable workflows
|
||||
1. **Noctalia theme change**: edit generated palette/theme source before hardcoding per-app colors; terminal text color must match accent in both palette and terminal override. `noctalia msg` for IPC; `git add && git commit && git push` in dotfiles repo.
|
||||
2. **Niri display rotation**: `niri msg outputs` → read `display.kdl` → edit transforms → `niri validate -c <path>` → `niri msg action load-config-file` → verify → git commit/push.
|
||||
3. **Ghostty AA fix**: replace `font-hinting = hinting` with `freetype-load-flags = hinting,light`. Reload via USR1: `pgrep ghostty | while read pid; do kill -USR1 "$pid"; done`.
|
||||
4. **Handy restart**: `pkill -f handy`, then hub daemon start with pty=true, persist=true, ready timeout 10s.
|
||||
5. **llama.cpp parallelism**: layer (default) = pipeline; tensor (experimental) = true tensor parallelism requiring flash attention + non-quantized KV; row (deprecated); none = single GPU. `--parallel` is serving-level concurrency, not model parallelism. No context/sequence parallelism mode.
|
||||
|
||||
## Pitfalls
|
||||
- Portrait 180° rotation in niri is 90°↔270° swap, NOT adding 180°.
|
||||
- Tensor mode: flash attention required, KV cache non-quantized, MoE/Mamba/SSM unsupported, `--fit` disabled.
|
||||
- Authenik session claim caching causes silent 403s post-group-add.
|
||||
- Local dir `serverAppConfig` vs remote repo `serverAppConfigs` name mismatch (trivial, pull/push unaffected).
|
||||
- Git LFS lock warnings: mitigate with `git config lfs.http://<repo>/info/lfs.locksverify true`.
|
||||
- Noctalia crispness: no single compositor-AA switch; scoped options only (hard corners; hard corners + no blur/noise; full crispness with opacity 1.0/blur 0/fontconfig AA off; terminal-only).
|
||||
@@ -0,0 +1,25 @@
|
||||
- Noctalia v5 config layering (verified live on blankpad 2026-10-09): bar widget layout lives in the GUI-override layer ~/.local/state/noctalia/settings.toml under [bar.name] keys start/center/end (defaults: start=["launcher","wallpaper","workspaces"], center=["clock"]); settings.toml loads LAST and wins over ~/.config/noctalia/config.toml. Both layers are watched and hot-reloaded — edit + noctalia config validate + screenshot is the full apply/verify loop (no restart, no reload IPC). Noctalia writes THROUGH symlinks and keeps the link, so keeping settings.toml symlinked from the command-center dotfiles repo is safe. Combine workspace+window indicator: [widget.taskbar] group_by_workspace=true (needs compositor per-window workspace data — niri has it) renders per-workspace capsules with disc labels, workspace_group_content = icons|count|dots (count/dots left-click activates the workspace); the workspaces widget itself has no window-display mode. Applied on blankpad: bar.default start=["launcher","wallpaper","workspaces","taskbar"], center=["cat"] (command-center commits 9255a8b, 0ccdd46). _(context: Blankpad noctalia 5.2.1 bar layout change 2026-10-09: taskbar moved to left section, group_by_workspace identified as the combining widget, verified via config validate + fullscreen screenshot)_
|
||||
- The command-center (~/.omp) pre-commit secret gate (.githooks/pre-commit, single grep -E of known token regexes) SELF-MATCHES its own documented pattern literals: committing prose that quotes them verbatim (e.g. a quoted PEM private-key header line, the xox[baprs] family prefix, the github-pat prefix, long sk- examples) trips SECRET GATE and blocks the commit (observed twice 2026-10-09 on agent/memories/--home-blank-.omp--/learned.md audit prose). Convention: redact pattern literals descriptively in prose ("PEM private-key headers", "the xox[baprs] family", "ghp 36") instead of setting ALLOW_SECRETS=1; the gate firing on prose is the gate working as designed. Note the gate covers only 10 families — no hf_/npm_/AIza/glpat-short/basic-auth-URL coverage — so clean-gate history is not proof of no leaks.
|
||||
- niri 26.04 event-stream (niri msg --json event-stream) has NO hotplug event: OutputConfigChanged is only an ACK that an IPC/config output change was Applied (per niri-ipc docs; proven live — mode flips via niri msg output NAME mode WxH@Hz and DPMS off/on emit zero output events). Detect monitor connect/disconnect by diffing the sorted unique set of .WorkspacesChanged.workspaces[].output across events (off/on cycle of eDP-1 emitted exactly WorkspacesChanged 2x). Debounce ~1.5s; OutputConfigChanged events are emitted in bursts on every config reload — ignoring them prevents hotplug-daemon echo loops. Implemented in dank/command-center scripts/desk-autoswitch.sh (blankpad-desk-autoswitch.service). Related trap: niri msg output 26.04 syntax is niri msg output NAME ACTION [args] (mode/off/on), not the old NAME WxH@Hz form.
|
||||
- On Frank's workstations (Arch, pam_faillock deny=3), subagents MUST NEVER run sudo/yay/pacman: in the harness sudo has no TTY, so every attempt logs 3 pam_unix failures and trips pam_faillock "temporarily locked out" (~600s), which re-arms on every retry and locks out the USER's real interactive sudo — observed 2026-10-09 on blankpad when a Handy-install subagent looped sudo yay -Syu handy-bin for 15 min; the user saw "incorrect password"/lockout and (understandably) suspected his password was changed. Diagnostics that settle it in one step: stat /etc/shadow (unchanged mtime = no password change), journalctl | grep 'sudo\[' (PWD/COMMAND identify the spammer), faillock --user blank. Recovery: kill the loop, then faillock --user self --reset works unprivileged (per-user dir /run/faillock/user is owner-writable); then user runs sudo -v. Any package install requiring root must be handed to the user as exact copy-paste commands, never attempted by an agent.
|
||||
- Tool traps verified on blankpad 2026-10-09 (niri/noctalia/Handy DE session): (1) niri msg action (and other niri msg subcommands that print help/action lists) write to STDERR — niri msg action 2/dev/null | grep ... yields empty output and looks like "feature missing"; always pipe with 2&1. Confirmed live: niri 26.04 has swap-window-left/right (window walks sideways, entering/leaving columns; column order untouched), move-window-to-monitor-*, focus-window-up/down (tabbed columns navigate with up/down; focus-window-left/right does NOT exist — no gap to fill). (2) frank-zealot (llamacpp dankclaw-3090 :7771) as a subagent model crashed mid-run with invalid string: surrogate U+D800..U+DBFF when a tool-call JSON argument contained many emoji — server-side JSON parse, type=server_error; keep subagent prompt/context payloads ASCII (no emoji) when agents run on fleet llama.cpp/radiance models; the crash is infrastructure, retry after stripping emoji. (3) The failed agent HAD already written its files before crashing (omp-docs/tools/handy.md, noctalia-shell.md rewrite survived; only its cleanup grep died) — after a subagent exit-1, always inspect target files for partial/complete output before respawning; its summary claims are unverified, diff against live truth (it invented a Handy settings JSON shape and flipped post_process_enabled true; live value false, real shape has bindings.transcribe/current_binding). (4) edit tool payload format: every change block MUST start with a PUT N.=M: (or DEL N.=M) hunk header followed by +/- lines; bare N:text + +text lines are rejected with "payload line has no preceding hunk header". (5) blankpad repo fact for pending work: live ~/.config/niri/cfg/{misc,rules}.kdl symlinks point at the GENERIC dotfiles/niri copies, which pin render-node /dev/dri/renderD129 "Intel Arc" — blankpad is single AMD Cezanne (renderD128); the newer dotfiles/zz-blankpad copies (no render-node, cursor 36, blur off, noctalia layer rules) are the correct live files, _(context: Grilling session on blankpad niri 26.04 / noctalia 5.2.1 / Handy 0.9.6 DE config; fact-check sweep + doc rewrites in dank/command-center omp-docs.)_
|
||||
- Blankpad DE stack, verified 2026-10-09 against live system + upstream (source of truth for future niri/noctalia/handy work; supersedes omp-docs claims): (1) Noctalia is v5 (installed 5.2.1-1.1, native noctalia binary). The v4 Quickshell line is DEAD upstream — repo doc omp-docs/tools/noctalia-shell.md (v4.7.7, qs -c noctalia-shell ipc call …) is stale; correct API is noctalia msg cmd, full list from noctalia msg --help (~110 cmds: screenshot-region/-fullscreen/-annotate, window-switcher, clipboard-*, workspace-alert-*, theme-mode-toggle, session lock-and-suspend, dpms-*, greeter-sync, templates-apply). Noctalia's niri docs (docs.noctalia.dev/noctalia/compositor-settings/niri/): autostart spawn-at-startup "noctalia", lid policy via niri switch-events { lid-close { spawn "noctalia" "msg" "session" "lock-and-suspend"; } }, wallpaper/blur layer-rules, [shell] niri_overview_type_to_launch_enabled. (2) niri 26.04 IS latest (repo moved to niri-wm org). New since docs: mainline blur/xray (background-effect, popups block, layer matcher), include optional=true + ~ expansion, niri msg action load-config-file --path file (live-swap config path — key for dotfiles symlink testing), cast IPC (niri msg casts, stop-cast), screenshot-window show-pointer=true, output options variable-refresh-rate (+on-demand=true), focus-at-startup, backdrop-color, per-output layout{}, mode custom=true/modeline (25.11), Escape cancels drag-and-drop. (3) Handy (cjpais/Handy) 0.9.6 installed via local/handy-bin; latest 0.9.8 (2026-10-03). Upstream LAW: SIGUSR1 handler REMOVED (WebKitGTK owns SIGUSR1; pkill -USR1 handy can crash the app) — toggle is SIGUSR2 or CLI flags handy --toggle-transcription / --toggle-post-process / --cancel; startup flags --start-hidden --no-tray. Wayland: compositor must own global binds (call the CLI flags from niri binds); text input needs wtype/dotool (NOT installed on blankpad; current paste_method=none + copy_to_clipboard, so no tool needed); overlay disabled-by-default up _(context: Fact-check sweep of niri/noctalia/handy docs on blankpad, 2026-10-09, ahead of a pending config-change grilling (Round 1 questions pending user answers))_
|
||||
- command-center (~/.omp) secret hygiene, audited 2026-10-09 at 79e780f: repo history is CLEAN — no live value (gitea/hf_token/hf_token_write/kagi_api_key) in any of 77 revs / 765 reachable blobs / 2 dangling blobs / 23 git-lfs payloads; the only sk-shaped string is a 35-byte unreachable fixture whose body is the alphabet + 1234. DURABLE GAP: .githooks/pre-commit (core.hooksPath=.githooks) matches only the OpenAI-style 18+ alnum token, ghp 36, github-pat 22+, glpat 20, the xox[baprs] family, AWS key ids, PEM private-key headers, aes 40-char, eyJ….eyJ (JWT). It does NOT match hf 30, npm 30, pypi, AIza 35, Stripe live/test keys, Telegram bot tokens, or a scheme://user:token@host basic-auth URL — the basic-auth shape is exactly what sat in remote.origin.url on both workstation clones until 2026-10-09, so the gate would never have caught it. The gate DID fire correctly on a synthetic token literal quoted in a commit (2026-10-09) AND on the audit prose itself (the gate self-matches its own documented pattern literals — quote them descriptively in docs/memory), so redact literals in docs/commit messages instead of using ALLOW_SECRETS. Live secrets live in ~/.config/secrets/* (dir 0700, files 0600 — hf_token_write was 0644 and was fixed); ~/.cache/huggingface/token duplicates hf_token at 0600 by HF convention.
|
||||
- Gitea client auth on Frank's workstations (2026-10-09): the HTTP token lives ONLY in ~/.config/secrets/gitea (0600) behind a global helper git config --global credential.helper 'store --file ~/.config/secrets/gitea'. All remote URLs are bare (http://192.168.0.143:3000/dank/repo.git) — a token embedded in remote.origin.url leaks via .git/config, git remote -v, and the git-lfs "Locking support detected on remote" warning, which echoes the full URL on every push (silence it per repo with git config lfs.bare-url/info/lfs.locksverify true). TRAP, measured on git 2.56.0: git credential-store does NOT parse the protocol=/host=/username=/password= block shape in the store file — that shape is only what git credential fill accepts on stdin, and the man page's ~/.git-credentials example is misleading. A block-format file yields a SILENT empty get, git falls through to askpass and dies with "could not read Username for …" (SSH_ASKPASS=/usr/bin/false in the omp harness env). The store file must be one URL per line WITH the port: http://dank:token@192.168.0.143:3000. Verify before trusting it, with prompts disabled so failure is loud: printf 'protocol=http\nhost=192.168.0.143:3000\n\n' | git credential-store --file ~/.config/secrets/gitea get must print username+password; then GIT_TERMINAL_PROMPT=0 git ls-remote --heads origin main and git push --dry-run. Rotation = rewrite that one file, nothing else. Documented in omp-docs/tools/gitea/client-access.md (command-center commit cfed6ef).
|
||||
- The origin remote of the command-center repo (~/.omp) embeds the Dankclaw Gitea token in plaintext: http://dank:token@192.168.0.143:3000/dank/command-center.git in .git/config. Consequence observed 2026-10-09: git push origin main echoes that URL (the git-lfs "Locking support detected" hint prints the full URL with the token) into stdout, so any transcript/log of a push contains a live credential. Recommended remediation proposed to Frank and NOT yet approved/applied — do not change remote auth unilaterally: git remote set-url origin http://192.168.0.143:3000/dank/command-center.git plus git config credential.helper 'store --file ~/.config/secrets/gitea' (that directory is Frank's existing 600-perm secret convention), and git config lfs.http://192.168.0.143:3000/dank/command-center.git/info/lfs.locksverify true to silence the lock hint. Reuse rule: when a token must be read from this repo, read it from the existing origin URL and reference its location only; never copy it into docs, skills, memory, or commit messages. Repo hooks: core.hooksPath=.githooks → pre-commit secret gate (patterns incl. eyJ… JWT shapes; override ALLOW_SECRETS=1) and pre-push git lfs pre-push, so pushes carry LFS objects.
|
||||
- Fleet LLM model ids in ~/.omp/agent/models.yml were renamed to the unit-name scheme on 2026-10-08 (committed 0cb021b, pushed 2026-10-09): frank-medium → frank-archon (llamacpp dankserver, 192.168.0.235:7770, ctx 200000 / maxTokens 16384), frank-large → frank-dragoon (dankclaw, 192.168.0.143:7770, 217088 / 45056, text+image), frank-small → frank-zealot (dankclaw-3090, 192.168.0.143:7771, 200704 / 16384). Service/unit/script names are UNCHANGED — /v1/models still advertises frank-primary, frank-large, frank-secondary. LAW (verified live with chat-completion probes on all three): radiance, halogen-flash-server and llama.cpp all accept an arbitrary request model id, so a harness-side id rename needs no endpoint change; the response model field just echoes the served name. Server-side compaction moved from dankserver:7770 to 192.168.0.143:7771 (frank-zealot), thresholdPercent 70 → 85, and every model's compactionModel now points at llamacpp (dankclaw-3090)/frank-zealot so the wire id stays valid at the compaction endpoint. Provider names in models.yml are labels referenced from config.yml (modelRoles, task.agentModelOverrides) and from compactionModel — renaming one requires updating every reference; verify by parsing both YAML files and cross-checking all llamacpp (host)/id tokens (regex must allow the space+parens in the provider label; a naive whitespace split yields false "unresolved" hits). Also: git renders whole-directory renames as mass delete + mass untracked — the 2026-10-09 tree showed 116 D + 73 ?? for the managed-skills consolidation (94 → 25) and the memory-skill re-bucketing (dotfile-git-workflow → ghostty-text-aa + niri-display-rotation + noctalia-theme; handy-service-mgmt → handy-flm-inference; inference-runtime → llama-cpp-parallelism). Nothing was lost: index matched HEAD, so git restore recovers any of it.
|
||||
- Measuring WHICH card is the bottleneck in a llama.cpp layer-split LLM server on a fleet host (frank-medium, Dankserver 192.168.0.235, 2x R9700 asymmetric PCIe: ROCm0=x16 ~31.5GB/s, ROCm1=x1 ~1.97GB/s) — two traps that produced wrong numbers on 2026-09-18: (1) RUN THE TRANSIENT MEASUREMENT SERVER ON AN ISOLATED PORT, not the production port. If it takes over the prod port (7770), live fleet traffic from other agents hits it and interleaves with your timed request — I measured 35:30 at 2.51 t/s on :7770 vs the true 31.5 t/s; on isolated :7790 (prod stopped to free VRAM, fleet hits the down 7770) it read 31.52 t/s. Override LISTEN_PORT in a temp env copy (sed), point curl at the new port. (2) Use AVERAGE rocm-smi --showuse over a LONG request (2k tokens, ~60s), sampled every 0.5s — PEAK util lies for a layer split because both cards are touched every token and both hit ~100% at peak. Average is the bottleneck indicator. FINDING (durable, specific to frank-medium): the x1 card (ROCm1) is the STRUCTURAL bottleneck at 98% avg util in every config — 35:30/36:29, drafter on ROCm1/ROCm0 all gave util0(x16)=41-43% / util1(x1)=98%, decode ~31.5 t/s in all four. It's COMPUTE-bound, not PCIe-bound (only ~5.2 MB activation crosses the x1 link per 512-token ubatch). The per-step cost is the verify phase (both cards run the full target) + the LOGITS, which are pinned to the card holding the last layers; the MTP drafter is a small nextn head so moving it (SPEC_DRAFT_DEVICE=ROCm0) buys nothing. 35:30 is the optimal split (balanced VRAM 1.27/1.85 GiB free); 36:29 buys 2 util points on the x16 card and erodes its VRAM (fit free 0.45 GiB) for no throughput gain. No llama-server flag moves the lm_head off the last-layer card, so the x1 bottleneck is architectural, not a mis-tune. Recorded in scripts/frank-medium.env (committed 6c6464a).
|
||||
- placeholder4
|
||||
- placeholder3
|
||||
- placeholder2
|
||||
- placeholder
|
||||
- frank-medium (Dankserver 192.168.0.235, llama.cpp layer-split 2x R9700 gfx1201) split-balance findings, 2026-09-17: (1) The LAYER SPLIT is NOT a decode-throughput lever at the 393216/MTP shape — 34:31, 35:30 and 36:29 all land in the same ~32 tok/s band (36:29 measured 31.76; 35:30 ~32.4 per env); total work, not imbalance, is the limit. (2) The x1 card (ROCm1, 0b:00.0, Gen4 x1 via B650 chipset ~1.97 GB/s) runs HOT (util ~91-97%) at 35:30 while the x16 card (ROCm0, 03:00.0, Gen4 x16 ~31.5 GB/s) idles (~45%) — because the MTP drafter + output layers + heavier compute buffers are STRUCTURALLY pinned to the card holding the LAST layers, NOT because the PCIe link is the bottleneck (only ~5.2 MB of boundary activation crosses per 512-token ubatch). (3) 36:29 makes the x16 card the utilized worker (util0=42%, util1=3%) at no throughput cost, but spends VRAM headroom (card0 free 0.45 GiB, under the 1.0 GiB floor; 2.22 GiB gap). Frank's standing preference: the card with the FASTEST PCIe lane (x16 = ROCm0) should be the more-utilized card. Deploying 36:29 requires re-fitting the seam test (tests/frank_medium_unit.bats) slack/floor, which currently rejects it.
|
||||
- niri output config on blankpad (niri 26.04): transform values are COUNTER-CLOCKWISE, so "90" means the panel's top bezel must physically sit on the LEFT ("270" = top bezel right). Two traps when pinning outputs in ~/.config/niri/cfg/display.kdl (symlink into ~/.omp/dotfiles/zz-blankpad/): (1) an output "X" {} block RESETS every property you don't restate to its default — eDP-1 needed scale 1.25 repeated or it would silently drop to 1.0; (2) niri re-runs its auto-layout for any output without a pinned position, so pinning only the externals shoved eDP-1 to the far right (3936,0) — pin eDP-1 at x=0 too if you want the original left-to-right order. Apply with niri validate -c ~/.config/niri/config.kdl then niri msg action load-config-file; confirm with niri msg outputs (shows Logical size swapped to 1200x1920 and "Transform: 90° counter-clockwise"). Live one-off rotation without config: niri msg output NAME transform 90. _(context: Rotated both external PM1601 panels (DP-2, HDMI-A-1) to portrait on blankpad 2026-09-15; verified via niri msg outputs, committed 2baf0c0 to dank/command-center.)_
|
||||
- Frank's frank-large ceiling is now 200,000 context in ~/.omp/agent/models.yml (provider llamacpp (dankclaw), committed 60dc0bb to dank/command-center 2026-09-13), while halogen on 192.168.0.143:7770 still serves context=262144 — a deliberate under-serve, not drift. Convention when changing any contextWindow there: read the served value from curl baseUrl/health (it reports context, model, vision.enabled) and check KV-pool admission before/after — admission reserves prompt + max_tokens against the pool (401,408 positions since 2026-09-13), so contextWindow + maxTokens must stay under it or the request hangs the full 3600 s queue timeout before a 503. maxTokens stays pinned at 65536 because HALOGEN_MAX_TOKENS_CAP 400s anything larger. Also: OMP reads models.yml only at startup, so a running session keeps the old ceiling; tell the user new omp invocations pick up the change.
|
||||
- Frank's terminal font stack on the workstation: package ttf-terminus-nerd (Arch, group nerd-fonts), which provides the TERMINESS outline derivative — families "Terminess Nerd Font" / " Mono" / " Propo". The bitmap original terminus-font is NOT installed, so classic-Terminus crispness depends entirely on 1-bit hinted rendering. Crisp/bitmap look = freetype-load-flags = hinting,monochrome (monochrome = 1-bit, disables AA; the light flag stays in the resolved value but is inert once monochrome is on, per ghostty.5.md). Sizing rule that matters: pixel cell = pt × (96/72) × monitor_scale, and the crisp sweet spot is when that lands on Terminus' native bitmap heights (8x16, 9x18, 10x20, 11x22, 12x24). On Frank's eDP-1 (1920x1080 @ scale 1.25) font-size 12 → 20px → exactly the 10x20 grid; 14pt → 23.3px, fractional and blurry regardless of flags. Check scale with niri msg outputs. Also: ghostty +show-config prints only values differing from default, and Ghostty's default font-size IS 12 — so a 12pt setting vanishes from the diff output; confirm against +show-config --default before concluding the key was ignored.
|
||||
- NEVER send SIGUSR1 to ghostty on Frank's workstations — it does not reload the config, it kills/respawns the terminal (observed 2026-09-13: PID churn, Frank said "stop killing the terminal"). The old recipe in agent/memories/--home-blank-.omp--/MEMORY.md and skills/dotfiles_git/scripts/reload_ghostty.sh ("pgrep ghostty | while read pid; do kill -USR1 $pid; done") is WRONG for ghostty 1.3.1. Correct way to apply a changed ghostty theme/config without losing shells: press the default keybind ctrl+shift+, (action reload_config) in each window — confirmed as the default in ghostty +show-config --default. For applying theme changes desktop-wide without touching terminals: noctalia msg templates-apply regenerates rendered theme files, and niri msg action load-config-file reloads the compositor (reload-config is NOT a valid niri subcommand). Custom noctalia palettes live at ~/.config/noctalia/palettes/Name.json and are selected with noctalia msg color-scheme-set custom "Name"; that source is immune to the community-palette catalog/md5 sync, unlike editing the cache in ~/.local/state/noctalia/community-palettes/. Also: noctalia msg screenshot-fullscreen saves to ~/Pictures (no need to scrape the PNG out of the clipboard with wl-paste).
|
||||
- In OMP's ~/.omp/agent/models.yml there are NO per-model sampler keys. Sampling params (temperature, topP, topK, minP, presencePenalty, repetitionPenalty) are GLOBAL settings in config.yml, each defaulting to -1 = "do not send". The only per-model hook is compat.extraBody — a Recordstring, unknown of extra top-level fields merged into every request body. It is applied LAST in the openai-completions builder (applyOpenAIExtraBody runs after applyChatCompletionsCompatPolicy, via Object.assign), so extraBody wins over anything the compat policy or global settings set. Use wire names (top_p, top_k, min_p, presence_penalty, repetition_penalty), not OMP's camelCase setting names. Two traps that cost real debugging time: 1. preserve_thinking is NOT user-settable in models.yml. qwenPreserveThinking is auto-derived inside buildOpenAICompat from the AUTO-DETECTED thinkingFormat plus a local (loopback/RFC1918) baseUrl — and that derivation happens BEFORE applyCompatOverrides. So writing compat.thinkingFormat: qwen in models.yml does NOT turn on preserve_thinking. For a custom-named model (e.g. frank-large, which matches no Qwen id pattern) the only path is extraBody: { preserve_thinking: true }. 2. NEVER put chat_template_kwargs inside extraBody. Object.assign is shallow, so it replaces the whole chat_template_kwargs object and silently drops the reasoning_effort kwarg the compat policy already put there. Top-level preserve_thinking is honoured by halogen-flash-server (its own 400 message says the field "may also be sent as top-level fields"), so top-level is both sufficient and safe. Why it mattered: halogen-flash-server has no server-side sampler defaults — temperature absent or 0 is greedy decode. OMP sent nothing (global -1), so the endpoint was decoding greedy while appearing configured. _(context: Verified 2026-09-13 on the Dank fleet while pinning Qwen3.8-Flash-Next recommended sampling (temp 1.0 / top_p 0.95 / top_k 20 / min_p 0.0 / presence_penalty 0.0 / repetition_penalty 1.0) plus preserve_thinking onto frank-large. Ground truth read from the installed bundle: pi-ai/src/providers/openai-completions.ts (applyOpenAIExtraBody applied last), pi-ai/src/providers/openai-shared.ts (applyChatC)_
|
||||
- Frank's Gitea lives on Dankclaw at http://192.168.0.143:3000, single org/user dank. Repo inventory (verified 2026-09-12 via /api/v1/user/repos): agent (default master), command-center (main — this is the ~/.omp dotfiles/KB repo), frank-llm-bench (main), serverAppConfigs (main), vllmScripts (main). LESSON — never clone from a repo name Frank states verbally; resolve the exact full_name via the Gitea API first. He said "serverAppConfig"; the real repo is serverAppConfigs (plural). Cloning the guessed name would 404 with no useful signal. Working procedure: read the dank HTTP token from the existing origin URL in ~/.omp (git remote -v) — do NOT copy that token into memory, docs, or skills; reference its location only. Then GET /api/v1/user/repos and match on the real name/case/plurality before cloning. Local-vs-remote naming: the clone was placed at ~/Projects/serverAppConfig (singular, as Frank explicitly requested) while origin points at the plural repo. The path mismatch is intentional, not drift — pull/push work normally. Repo layout: dankclaw/, dankserver/, frankbook-pro/ (7 quadlet units: sonarr, radarr, lidarr, readarr, prowlarr, qbittorrent, gluetun), shared/ (homepage, alloy, loki, prometheus, searxng), tests/ (47-file E2E suite), serverAppConfigs-docs/ (115-file KB — read its index.md before touching the stack). _(context: Frank asked to "pull my serverAppConfig repository into ~/Projects/serverAppConfig"; Gitea API revealed the plural name. Credential handling: token reused from existing ~/.omp remote, deliberately not persisted here.)_
|
||||
- Ghostty 1.3.1 (and the 1.3.x line) has NO font-hinting and NO font-subpixel-render config key. The only font-smoothing key is freetype-load-flags, a comma-separated list: hinting, force-autohint, monochrome, autohint, light (each default-enabled except monochrome/force-autohint; prefix no- to disable). Ground truth: /usr/share/ghostty/doc/ghostty.5.md:598-628. To actually DISABLE anti-aliasing you need freetype-load-flags = hinting,monochrome (docs line 617: monochrome = 1-bit rendering, disables AA). Note hinting,light is a NO-OP — both are already Ghostty defaults. Verification workflow before editing any ghostty key: ghostty +validate-config (exit 0 = valid) plus grep the manpage for the key name; an unknown key silently does nothing useful and +validate-config is the authoritative gate. This matters because a poisoned memory claim about font-subpixel-render/font-hinting caused 3 wasted commits (4c84f02, df33310, 268dcbd) editing a nonexistent key on 2026-09-11. _(context: Verified 2026-09-12 against ghostty 1.3.1-arch2.2 manpage and ghostty +validate-config on the workstation. Corrects contradictory entries in agent/memories/--home-blank-.omp--/learned.md (line 2 asserts font-subpixel-render is the key; it is not).)_
|
||||
- Ghostty's font antialiasing is controlled by font-hinting = no-hinting in the config file, not font-subpixel-render = off as the Noctalia reference skill states. When editing ghostty config, check /usr/share/ghostty/doc/ghostty.5.md or src/config/Config.zig — ghostty's --help confirms this. Invalid field causes font-subpixel-render: unknown field on restart. _(context: Discovered when user reported "font-subpixel-render: unknown field" after my initial edit to ghostty config)_
|
||||
- Noctalia is a Niri-based desktop environment at /usr/bin/noctalia. Config: ~/.config/noctalia/config.toml. Provides IPC via 'noctalia msg', launcher, lock screen, session menu, and 19 app templates (GTK, Ghostty, Alacritty, Hyprland, Sway, WezTerm, Helix, Btop, Starship, Qt, KDE, Cava, scroll, mango, umbriel, foot, emacs, kitty, alacritty). Templates under /usr/share/noctalia/assets/templates/. Color palette generator: 'noctalia theme [image]'. Text antialiasing: Ghostty uses 'font-subpixel-render' config key (set to 'off' to disable). Alacritty uses 'font.antialias = false' in alacritty.toml.
|
||||
@@ -0,0 +1 @@
|
||||
Frank workstation (user blank) facts and reusable workflows: Noctalia/Niri DE config and display rotation (portrait 180° = 90°↔270° swap), Ghostty/Alacritty text-AA fixes, niri compositor-AA limits, handy/FLM inference constraints, llama.cpp split modes, and dotfile git workflow. Dotfiles repo is ~/.omp (Gitea dank/command-center). Current layout: DP-2@0 → HDMI-A-1@1200 → eDP-1@2400. Repository state wins over this memory.
|
||||
@@ -0,0 +1,131 @@
|
||||
# Raw Memories
|
||||
|
||||
## 01a0a7ff-dae2-71fd-9bd0-a8aa7c9d3676
|
||||
updated_at: 1789533306
|
||||
## llama.cpp Multi-GPU Parallelism (verified against docs/multi-gpu.md on master)
|
||||
|
||||
### Split modes (--split-mode / -sm)
|
||||
- **layer** (default): Pipeline parallelism. Each GPU holds a contiguous slice of layers. Token flows through GPUs sequentially. Minimizes inter-GPU data movement; tolerates slow interconnects (PCIe fine). Best for prefill and fitting large models. Weakness: needs many tokens to keep pipeline full, poor low-batch latency.
|
||||
- **tensor** (experimental): True tensor parallelism. Splits each layer's weights AND KV cache across all GPUs via a 'meta device', doing cross-GPU all-reduce per layer (NCCL). Minimizes latency, parallelizes any workload size. Heavily bottlenecked by interconnect speed (wants NVLink + NCCL). Best for fast token generation.
|
||||
- **row** (deprecated): Row-split dense weights only. Predecessor to tensor; comparatively poor performance.
|
||||
- **none**: Single GPU, selected via --main-gpu.
|
||||
|
||||
### Key constraints on tensor mode
|
||||
- Requires flash attention ON (-fa on)
|
||||
- KV cache must be non-quantized: f32/f16/bf16 only
|
||||
- NOT implemented for: MoE/hybrid, Mamba/SSM, and several other architectures (hard-fails with SPLIT_MODE_TENSOR not implemented)
|
||||
- --fit is disabled; must size --ctx-size / --parallel manually
|
||||
|
||||
### Separate axis (NOT model parallelism)
|
||||
- **--parallel / -np**: Number of concurrent server slots (each with own KV cache). Serving-level concurrency/batch parallelism. On single GPU this is the main throughput knob.
|
||||
- **CPU+GPU offload** (--n-gpu-layers, --n-cpu-moe): Hybrid inference when model exceeds VRAM. Device-distribution strategy, not 'parallelism' in the strict sense.
|
||||
|
||||
### What does NOT exist in llama.cpp
|
||||
- No first-class context/sequence parallelism mode (unlike vLLM/TensorRT-LLM). If you see that term applied to llama.cpp, it's loose third-party language.
|
||||
|
||||
### User hardware context
|
||||
- Single RTX 4070 Max-Q (~12 GB VRAM). Multi-GPU split modes (layer/tensor) do not apply. Relevant knobs: --parallel for concurrency, CPU+GPU offloading for models >12 GB.
|
||||
|
||||
## 01a0a817-7382-7525-92f1-14cd7fade072
|
||||
updated_at: 1789529008
|
||||
## Blankpad niri layout: externals left of eDP-1
|
||||
|
||||
Final committed layout (83c9c92, 2026-09-16):
|
||||
- DP-2: x=0, portrait 1200x1920, transform 90 CCW, scale 1
|
||||
- HDMI-A-1: x=1200, portrait 1200x1920, transform 270 CCW, scale 1
|
||||
- eDP-1: x=2400, 1536x864, scale 1.25 (restated per block-reset trap)
|
||||
|
||||
X-offset chain: 0 → 1200 (DP-2 portrait width) → 2400 (HDMI-A-1 portrait width) → eDP-1.
|
||||
|
||||
Config path: ~/.omp/dotfiles/zz-blankpad/niri/.config/niri/cfg/display.kdl
|
||||
Repo: dank/command-center (main), local Gitea at 192.168.0.143:3000.
|
||||
|
||||
The skill niri-output-rotation-pinning was followed correctly end-to-end. No new pitfalls discovered beyond what the skill already documents.
|
||||
|
||||
## 01a0a798-f61e-71d3-8fb6-55879410c3fc
|
||||
updated_at: 1789518997
|
||||
## Niri Display Rotation (Portrait 180° Flip)
|
||||
|
||||
### Hardware
|
||||
- eDP-1: Lenovo laptop panel, 1920x1080, scale 1.25, normal transform
|
||||
- DP-2: PNP(XYM) PM1601, 1920x1200, scale 1, portrait (was 270° CCW → now 90° CCW)
|
||||
- HDMI-A-1: PNP(XYM) PM1601 demoset-1, 1920x1200, scale 1, portrait (was 90° CCW → now 270° CCW)
|
||||
|
||||
### Key Technical Facts
|
||||
- Niri config split into `~/.config/niri/cfg/*.kdl` (Noctalia-style includes)
|
||||
- Display config: `~/.config/niri/cfg/display.kdl` — per-output `transform` field in KDL
|
||||
- `niri msg outputs` — shows current transform, logical size, scale, position
|
||||
- `niri validate -c <path>` — dry-run validation before applying
|
||||
- `niri msg action load-config-file` — hot-reload config at runtime (no restart needed)
|
||||
- After rotation in portrait, logical size swaps (1920x1200 → 1200x1920) but logical position X may shift; verify with `niri msg outputs` after reload
|
||||
|
||||
### Critical Pitfall: Portrait 180° Rotation
|
||||
- In portrait, "flip 180°" = swap 90° ↔ 270° transform values
|
||||
- Setting `transform "180"` on a portrait monitor kicks it to landscape-upside-down
|
||||
- Niri transform values are absolute: `normal`, `90° counter-clockwise`, `180°`, `270° counter-clockwise` — no modular arithmetic
|
||||
|
||||
### Workflow (Reusable)
|
||||
1. `niri msg outputs` — capture current state
|
||||
2. Read `~/.config/niri/cfg/display.kdl`
|
||||
3. Edit transform values (90↔270 swap for 180° portrait flip)
|
||||
4. `niri validate -c ~/.config/niri/config.kdl` — confirm valid
|
||||
5. `niri msg action load-config-file` — apply
|
||||
6. `niri msg outputs | grep -E '^Output|Transform|Logical|Scale'` — verify
|
||||
7. `git add && git commit && git push` in dotfiles repo (`/home/blank/.omp`)
|
||||
|
||||
## 01a09e49-b17b-74c3-83e9-2eb95ed25621
|
||||
updated_at: 1789368866
|
||||
Task: user requested changing Noctalia theme yellow to red and terminal text color to same red; rollout ended before applying edits after an ask about crispness scope was interrupted.
|
||||
|
||||
Noctalia facts: Niri-based DE at /usr/bin/noctalia. Main config ~/.config/noctalia/config.toml. Niri config includes noctalia.kdl for display settings. IPC uses direct `noctalia msg <command>` instead of qs wrapper; key commands panel-toggle, session, lock, launcher. App templates in /usr/share/noctalia/assets/templates/ for gtk, ghostty, alacritty, hyprland, sway, wezterm, helix, btop, starship, qt, kde, cava, scroll, mango, umbriel, foot, emacs, kitty. Palette is generated from an image with `noctalia theme [image]`. Migration notes: Ghostty should use `theme = noctalia`; Alacritty can import noctalia.toml theme; GTK 3/4 accent palette default mentioned as #b267e4 purple; Niri keybinds should map to noctalia msg commands.
|
||||
|
||||
Theme-edit workflow: for changing a Noctalia accent such as yellow to red, prefer locating and editing the generated Noctalia palette/theme source before hardcoding per-app colors. If Ghostty is set to `theme = noctalia`, direct terminal color overrides may be less durable than updating the theme itself. Terminal text color should match the chosen red in both the Noctalia palette and any terminal-specific override if used.
|
||||
|
||||
Config discovery in this environment: expected ~/.config/noctalia/config.toml and ~/.config/ghostty/config were not successfully read due shaken/pruned tool output. Candidate locations included ~/.config/noctalia, ~/.config/ghostty, and /home/blank/.omp/dotfiles/noctalia/.local/share/noctalia under cwd /home/blank/.omp. Use ls/glob/read on specific files rather than assuming a single canonical path.
|
||||
|
||||
Niri crispness facts: in niri 26.04, true compositor AA is limited to corner rounding via niri_rounding_alpha() smoothstep and unused gradient/hue interpolation. Corner AA is driven by window-rule geometry-corner-radius with clip-to-geometry; setting radius 0 short-circuits the shader. Background blur/noise are not AA but cause softness: background-effect { blur true } plus top-level blur { passes 2, offset 3.0, noise 0.03 } creates frosted glass and film grain. Shadows were already off. Wayland clients rasterize text; niri does not alter glyph AA. Ghostty font AA is controlled by freetype-load-flags, while other apps use fontconfig or app-specific settings. No nearest-neighbor resample knob was found in this build.
|
||||
|
||||
Decision guidance: when asked to turn off all compositor AA for a crisp look, explain there is no single switch; offer scoped options: hard corners only; hard corners plus disable backdrop blur/noise; full maximum crispness with terminal opacity 1.0/blur 0 and global fontconfig antialias false; or terminal-only crispness. Warn that removing blur/noise changes Noctalia's glass aesthetic.
|
||||
|
||||
## 01a0993b-f83f-7331-9cad-fff3fc7f3070
|
||||
updated_at: 1789286338
|
||||
Target: agent/models.yml (frank-large entry). Sampler defaults overridden from halogen/greedy to temp=1.0, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0. Added preserve_thinking=true to retain thinking blocks across turns. Deployed to dank/command-center on dankclaw. Live effect triggers on next config read; no container restart required. Git LFS locking notice is advisory, does not block pushes. Repo path: ~/.omp.
|
||||
|
||||
## 01a096d7-c1b8-7291-b467-a8521fe9c3c6
|
||||
updated_at: 1789267061
|
||||
Noctalia DE (Niri-based): bin /usr/bin/noctalia; configs ~/.config/noctalia/config.toml + noctalia.kdl. IPC via noctalia msg <cmd> (replaces qs -c noctalia-shell ipc call). Key cmds: panel-toggle, session, lock, launcher. App templates /usr/share/noctalia/assets/templates/. Palette gen noctalia theme [image]. GTK accent #b267e4. Text AA: Ghostty font-subpixel-render=off, Alacritty font.antialias=false.
|
||||
FLM NPU (Dankclaw): gemma4-it:e4b, port 7771. Supports PNG/JPEG; rejects WebP (convert first). Requires thinkingLevelMap in models.json. FIFO queue, ~7-11s model switch. Pitfall: missing .xclbin symlinks under ~/.config/flm/xclbins/ cause failures.
|
||||
Frank desktop: cursor 36 (gsettings org.gnome.desktop.interface cursor-size, GTK settings.ini, XCURSOR_SIZE=36), theme capitaine-cursors. Authenik 2026.8 outpost caches claims per session (1-24h TTL), causes silent 403s post-group-add despite correct membership.
|
||||
Halogen runtime (frank-large): baked-in chat template, no file swap. Accepts chat_template_kwargs (enable_thinking, preserve_thinking, reasoning_effort). frank-small uses checkpoint's fixed Qwen3.6 template; deploy-plan.py mapping is vestigial.
|
||||
|
||||
## 01a09618-6a77-71d5-a5b2-36e82045b695
|
||||
updated_at: 1789227719
|
||||
OMP harness config/dotfiles repo located at ~/.omp.
|
||||
Cloned serverAppConfig to ~/Projects/serverAppConfig (HEAD 701ab14, main tracks origin/main, HTTP transport with embedded dank token, no SSH keys).
|
||||
Local dir is singular (serverAppConfig), remote repo is plural (serverAppConfigs).
|
||||
Structure (308 files, 4.2 MB):
|
||||
- dankclaw/: Dankclaw stack defs.
|
||||
- dankserver/: Inference backend config.
|
||||
- frankbook-pro/: 7 quadlet units (sonarr, radarr, lidarr, readarr, prowlarr, qbittorrent, gluetun).
|
||||
- shared/: Homepage (html/css/js + services.json), alloy, loki, prometheus, searxng.
|
||||
- tests/: 47 files, E2E suite.
|
||||
- serverAppConfigs-docs/: 115 files, Architecture KB (deployment flow, network/storage/ testing policy, container netns, Dankclaw service pages).
|
||||
- old/, report/, scripts/, .githooks/, AGENTS.md: legacy, reports, tooling, hooks.
|
||||
Recent commits: observability quadlet cutover (loki/alloy/comfyui images, grafana OAuth secret per-service env seeding, docs commit recording exposure posture + unconfigured state).
|
||||
Pitfall: Local dir name mismatch vs remote repo name (trivial, pull/push unaffected).
|
||||
|
||||
## 01a08f58-4d8f-7356-8bcc-63a767971a5a
|
||||
updated_at: 1789224571
|
||||
[GIT_WORKFLOW] State check: git status --short | git branch --show-current | git log --oneline HEAD..@{u}. Branch: main. Untracked dirs: .pre-pull-backup/, home/.
|
||||
[DOTFILES_STATE] Modified: alacritty.toml, config.ghostty, noctalia.css (gtk3/4), niri .kdl files. Shift to noctalia theme.
|
||||
[GHOSTTY_FIX] Config edit: replace font-hinting = hinting with freetype-load-flags = hinting,light (line 10).
|
||||
[COMMIT_PUSH] Pattern: git add <file> && git commit -m "<msg>" && git push. Remote LFS lock warning present; mitigate via git config lfs.http://<repo>/info/lfs.locksverify true.
|
||||
[RELOAD] Ghostty reload via USR1: pgrep ghostty | while read pid; do /usr/bin/kill -USR1 "$pid"; done.
|
||||
|
||||
## 01a08f59-61c3-7103-bd4f-18ea24369233
|
||||
updated_at: 1789111781
|
||||
Target: handy app. Binary: /usr/bin/handy. Kill: pkill -f handy. Start: hub op=start, application=/usr/bin/handy, pty=true, persist=true, ready={log:'.*', timeout:10}. Result: Persistent daemon, 69 catalog models seeded.
|
||||
|
||||
## 01a08f50-c405-7763-8a24-b52be5af422f
|
||||
updated_at: 1789110966
|
||||
Error: 'Platform error: permission denied opening 16 device node(s) under /dev/input' causing handy keyboard listener failure. Diagnosis: ls -la /dev/input confirms devices are root:input with crw-rw---- (660). groups shows user 'blank' missing 'input'. Resolution: sudo usermod -aG input blank. Pitfalls: 1. usermod without sudo fails to lock /etc/passwd; 2. sudo password entry in this CLI environment requires PTY for reliable input.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a08f50-c405-7763-8a24-b52be5af422f
|
||||
updated_at: 1789110966
|
||||
|
||||
Fix 'permission denied opening device node(s)' in /dev/input by adding user to the 'input' group. Root cause: /dev/input/* devices are owned by root:input (660 permissions); user was missing 'input' group. Resolution: sudo usermod -aG input $USER. Note: sudo password handling in CLI requires PTY, and usermod needs sudo to lock /etc/passwd.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a08f58-4d8f-7356-8bcc-63a767971a5a
|
||||
updated_at: 1789224571
|
||||
|
||||
Durable git workflow for dotfile repositories: state inspection (short status, unpushed check), targeted config edits (Ghostty font-hinting fix), commit/push with LFS lock awareness, and USR1 process reloads.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a08f59-61c3-7103-bd4f-18ea24369233
|
||||
updated_at: 1789111781
|
||||
|
||||
Restart 'handy' application (/usr/bin/handy). Kill via pkill -f handy, start via hub daemon (pty:true, persist:true, ready timeout 10s). Status: Up, 69 catalog models seeded.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a09618-6a77-71d5-a5b2-36e82045b695
|
||||
updated_at: 1789227719
|
||||
|
||||
OMP harness lives at ~/.omp. Cloned serverAppConfig to ~/Projects/serverAppConfig (remote serverAppConfigs, HTTP/dank token transport). 308 files across dankclaw/, dankserver/, frankbook-pro/ (7 quadlets), shared/, tests/ (47 E2E), serverAppConfigs-docs/ (115). Recent work: observability quadlet cutover (loki/alloy/comfyui, grafana OAuth env seeding).
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a096d7-c1b8-7291-b467-a8521fe9c3c6
|
||||
updated_at: 1789267061
|
||||
|
||||
Noctalia (Niri) DE config & IPC patterns, FLM NPU constraints & WebP pitfall, Frank's desktop settings & cursor sizing, Authenik session claim caching, Halogen runtime baked-in template handling.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0993b-f83f-7331-9cad-fff3fc7f3070
|
||||
updated_at: 1789286338
|
||||
|
||||
Pinned frank-large sampler settings (temp=1.0, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0) + preserve_thinking=true in agent/models.yml. Pushed to dank/command-center (Gitea, 192.168.0.143:3000). Applies on next config read without service restart.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a09e49-b17b-74c3-83e9-2eb95ed25621
|
||||
updated_at: 1789368866
|
||||
|
||||
Noctalia theme change request (yellow to red, terminal text same red) led to locating Noctalia/Niri config surfaces and crispness controls. No final edit was applied before abort. Durable: Noctalia is Niri-based; palette generated via noctalia theme image; Ghostty uses theme=noctalia; niri compositor AA only affects corner rounding, while blur/noise cause soft glass look and text AA is client-side.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0a798-f61e-71d3-8fb6-55879410c3fc
|
||||
updated_at: 1789518997
|
||||
|
||||
Rotated two external PNP(XYM) PM1601 portable monitors 180° in portrait via niri display config edit. Key insight: in portrait mode, a 180° flip is the 90°↔270° transform swap, not adding 180°. Workflow: niri msg outputs → edit cfg/display.kdl → niri validate → niri msg action load-config-file → verify → git commit/push.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0a7ff-dae2-71fd-9bd0-a8aa7c9d3676
|
||||
updated_at: 1789533306
|
||||
|
||||
llama.cpp parallelism types: two real model-parallelism modes (pipeline/layer split = default, tensor = experimental), plus deprecated 'row' and single-GPU 'none'. --parallel is a separate serving-level concurrency axis, not model splitting. No first-class context/sequence parallelism. Tensor mode requires flash attention, non-quantized KV, and excludes MoE/Mamba/SSM architectures.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0a817-7382-7525-92f1-14cd7fade072
|
||||
updated_at: 1789529008
|
||||
|
||||
Re-arranged blankpad niri display: moved two rotated PM1601 externals (DP-2, HDMI-A-1) from right of eDP-1 to left. New x-chain: DP-2@0 → HDMI-A-1@1200 → eDP-1@2400. Committed as 83c9c92 to dank/command-center.
|
||||
@@ -0,0 +1,13 @@
|
||||
# Ghostty Text AA Fix
|
||||
|
||||
Ghostty font AA is controlled by `freetype-load-flags`, NOT `font-hinting`.
|
||||
|
||||
## Fix
|
||||
- Replace `font-hinting = hinting` with `freetype-load-flags = hinting,light` in `~/.config/ghostty/config`.
|
||||
|
||||
## Reload
|
||||
- Send USR1 to all ghostty processes: `pgrep ghostty | while read pid; do kill -USR1 "$pid"; done`
|
||||
|
||||
## Related
|
||||
- Alacritty text AA: `font.antialias=false` in its config.
|
||||
- Noctalia crispness: niri compositor AA only affects corner rounding (`niri_rounding_alpha()`); text AA is client-side. No nearest-neighbor resample knob in this build.
|
||||
@@ -0,0 +1,17 @@
|
||||
# Handy + FLM NPU Inference
|
||||
|
||||
## Handy app
|
||||
- Binary: `/usr/bin/handy`
|
||||
- Start: `hub op=start` with `application=/usr/bin/handy`, `pty=true`, `persist=true`, `ready={log:'.*', timeout:10}`
|
||||
- Result: persistent daemon, 69 catalog models seeded
|
||||
- Restart: `pkill -f handy`, then re-run hub op=start
|
||||
|
||||
## FLM NPU (Dankclaw)
|
||||
- gemma4-it:e4b on port 7771
|
||||
- Accepts PNG/JPEG; rejects WebP (convert first with ImageMagick `convert`)
|
||||
- Requires `thinkingLevelMap` in models.json
|
||||
- FIFO queue; ~7-11s model switch
|
||||
- Pitfall: missing `.xclbin` symlinks under `~/.config/flm/xclbins/` cause failures
|
||||
|
||||
## Related
|
||||
- Dotfile repo: `~/.omp` (Gitea `dank/command-center`, 192.168.0.143:3000)
|
||||
@@ -0,0 +1,23 @@
|
||||
# llama.cpp Multi-GPU Parallelism
|
||||
|
||||
## Split modes (--split-mode / -sm)
|
||||
- **layer** (default): Pipeline parallelism. Each GPU holds a contiguous layer slice. Token flows sequentially through GPUs. Minimizes inter-GPU data movement; tolerates slow interconnects (PCIe fine). Best for prefill and fitting large models. Weakness: needs many tokens to keep pipeline full, poor low-batch latency.
|
||||
- **tensor** (experimental): True tensor parallelism. Splits each layer's weights AND KV cache across all GPUs via 'meta device', doing cross-GPU all-reduce per layer (NCCL). Minimizes latency, parallelizes any workload size. Heavily bottlenecked by interconnect speed (wants NVLink + NCCL). Best for fast token generation.
|
||||
- **row** (deprecated): Row-split dense weights only. Predecessor to tensor; comparatively poor performance.
|
||||
- **none**: Single GPU, selected via --main-gpu.
|
||||
|
||||
## Key constraints on tensor mode
|
||||
- Requires flash attention ON (-fa on)
|
||||
- KV cache must be non-quantized: f32/f16/bf16 only
|
||||
- NOT implemented for: MoE/hybrid, Mamba/SSM, and several other architectures (hard-fails with SPLIT_MODE_TENSOR not implemented)
|
||||
- --fit is disabled; must size --ctx-size / --parallel manually
|
||||
|
||||
## Separate axis (NOT model parallelism)
|
||||
- **--parallel / -np**: Number of concurrent server slots (each with own KV cache). Serving-level concurrency/batch parallelism. On single GPU this is the main throughput knob.
|
||||
- **CPU+GPU offload** (--n-gpu-layers, --n-cpu-moe): Hybrid inference when model exceeds VRAM. Device-distribution strategy, not 'parallelism' in the strict sense.
|
||||
|
||||
## What does NOT exist in llama.cpp
|
||||
- No first-class context/sequence parallelism mode (unlike vLLM/TensorRT-LLM). If that term is applied to llama.cpp, it's loose third-party language.
|
||||
|
||||
## Hardware context (this environment)
|
||||
- Single RTX 4070 Max-Q (~12 GB VRAM). Multi-GPU split modes do not apply. Relevant knobs: --parallel for concurrency, CPU+GPU offloading for models >12 GB.
|
||||
@@ -0,0 +1,26 @@
|
||||
# Niri Display Rotation Workflow
|
||||
|
||||
Used to rotate/flip external monitors in portrait mode.
|
||||
|
||||
## Workflow
|
||||
1. `niri msg outputs` — capture current transform, logical size, scale, position.
|
||||
2. Read `~/.config/niri/cfg/display.kdl`.
|
||||
3. Edit `transform` values.
|
||||
4. `niri validate -c ~/.config/niri/cfg/display.kdl` — dry-run validation.
|
||||
5. `niri msg action load-config-file` — hot-reload (no restart).
|
||||
6. `niri msg outputs | grep -E '^Output|Transform|Logical|Scale'` — verify.
|
||||
7. `git add && git commit && git push` in dotfiles repo (`~/.omp`).
|
||||
|
||||
## Critical pitfall
|
||||
- In portrait, "flip 180°" = swap `90° counter-clockwise` ↔ `270° counter-clockwise`.
|
||||
- Setting `transform "180"` on a portrait monitor kicks it to landscape-upside-down.
|
||||
- Niri transform values are absolute, NOT modular arithmetic.
|
||||
|
||||
## Verified layout (blankpad, 83c9c92)
|
||||
- DP-2: x=0, portrait 1200x1920, transform 90° CCW, scale 1
|
||||
- HDMI-A-1: x=1200, portrait 1200x1920, transform 270° CCW, scale 1
|
||||
- eDP-1: x=2400, 1536x864, scale 1.25
|
||||
|
||||
## Config split
|
||||
- `~/.config/niri/cfg/*.kdl` (Noctalia-style includes).
|
||||
- Display config lives in `~/.config/niri/cfg/display.kdl`.
|
||||
@@ -0,0 +1,21 @@
|
||||
# Noctalia Theme Editing
|
||||
|
||||
Noctalia is a Niri-based DE (`/usr/bin/noctalia`).
|
||||
|
||||
## Config locations
|
||||
- Main config: `~/.config/noctalia/config.toml`
|
||||
- Niri config: `~/.config/niri/cfg/noctalia.kdl`
|
||||
- IPC: `noctalia msg <command>` (direct, not qs wrapper). Key cmds: panel-toggle, session, lock, launcher.
|
||||
- App templates: `/usr/share/noctalia/assets/templates/` (gtk, ghostty, alacritty, wezterm, helix, btop, starship, qt, cava, scroll, mango, umbriel, foot, emacs, kitty, etc.)
|
||||
- Palette generation: `noctalia theme [image]`
|
||||
- GTK accent default: #b267e4
|
||||
|
||||
## Changing an accent (e.g. yellow → red)
|
||||
1. Locate and edit the generated Noctalia palette/theme source BEFORE hardcoding per-app colors.
|
||||
2. If Ghostty is set to `theme = noctalia`, direct terminal color overrides are less durable than updating the theme itself.
|
||||
3. Terminal text color must match the chosen accent in BOTH the Noctalia palette and any terminal-specific override.
|
||||
4. Hot-reload: `noctalia msg action load-config-file` (or equivalent config load action).
|
||||
|
||||
## Related
|
||||
- Niri display config: `~/.config/niri/cfg/display.kdl`.
|
||||
- Dotfile git workflow: `git add && git commit && git push` in `~/.omp`.
|
||||
@@ -0,0 +1,108 @@
|
||||
# Frank Fleet Memory Consolidation
|
||||
|
||||
## 1. Fleet Topology & Identity
|
||||
|
||||
### Hosts
|
||||
- **dankclaw** (192.168.0.143): Strix Halo APU (gfx1151, ~114 GiB GTT) + M.2 docked RTX 3090 (24 GiB, PCIe x4). Hosts `frank-large` (Halogen) and `frank-small` (Llama.cpp).
|
||||
- **dankserver** (192.168.0.235): ASUS B650-E, 2x AMD R9700 (gfx1201, 32 GiB each), 4x DIMMs DDR5. Hosts `frank-primary` (VLLM/Radiance) and `frank-medium` (Llama.cpp).
|
||||
- **frankbook-pro** (192.168.0.112): M-series Mac mini (~30 GiB RAM). Role unclear/secondary.
|
||||
|
||||
### Services (Live as of 2026-10-05)
|
||||
- **frank-large** (:7770): Halogen-flash-server v2, Qwen3.8-27B Uncensored (Q6_0/Q8_0/ROCmFP4), MTP/DFlash2 drafter, HALOGEN_CTX=262144, KV pool 524288 pos.
|
||||
- **frank-small** (:7771): Llama.cpp, MiMo-V2.6-9B (Q6_K/Q5_K_M) or Qwen3.8-9B (IQ3_S), 2x200704 f16-KV.
|
||||
- **frank-primary** (:7770): tcclaviger VLLM fork, Qwen3.8-Flash-Next FP8, TP=2, SEQS=3, MNBT=2048, MAXLEN=200704, Arena ~17.77 GiB.
|
||||
- **frank-medium** (:7770): Llama.cpp, Qwen3.8-27B UD-Q8_K_XL, Tensor Split 35:30, CACHE_RAM 16384 MiB.
|
||||
- **Authentik**: `ridley` shape template, Django ORM creation, Group-based ACLs.
|
||||
- **ComfyUI**: Strix Halo iGPU, FP8/Q6 quantization, Ghostty image protocol.
|
||||
|
||||
## 2. Hardware & Memory Constraints
|
||||
|
||||
### RAM Stability (dankserver)
|
||||
- **Critical**: 4x DIMMs at 4800 MT/s caused silent corruption (NaN/garbage) at ~40k tokens in `frank-primary`. **Fix**: Drop to 4600 MT/s.
|
||||
- **Lesson**: Consumer non-ECC RAM raises no MCE for silent flips. Short in-band checks miss intermittent corruption.
|
||||
- **Shape E**: 3.15e9 B KV pool, 17.77 GiB Arena, UTIL 0.982.
|
||||
|
||||
### GPU & PCIe
|
||||
- **dankclaw 3090**: PCIe x4 dock (~29 GB/s). Decode bottlenecked by graph splits + PCIe sync. 30 t/s unreachable without mainline PR #28901.
|
||||
- **dankserver R9700**: Chipset-attached PCIe (Gen4 x1) fails RCCL atomics. **Fix**: M.2→PCIe x4 riser for 2nd card.
|
||||
- **Strix Halo**: 64 GiB Unified, RADV Vulkan. HALOGEN_CTX must be 262144 (native Qwen3.8-Flash-Next).
|
||||
|
||||
### KV & Context
|
||||
- **frank-large**: `maxTokens` 57344 (harness) + HALOGEN_CTX 262144. 400 error if `prompt + max > ctx`.
|
||||
- **frank-medium**: `CACHE_RAM` 16384 MiB required for 196608 ctx DeltaNet state. 8192 MiB caused refusals.
|
||||
- **frank-small**: 2x200704 f16-KV slots.
|
||||
|
||||
## 3. Runtime Environments
|
||||
|
||||
### Halogen (dankclaw)
|
||||
- **Container**: `halogen-flash-server v2` (Swift1.5 Abliterated).
|
||||
- **Env**: `HALOGEN_CTX=262144`, `HALOGEN_KV_POOL_POSITIONS=524288`.
|
||||
- **Pitfalls**: `HALOGEN_FLASH_PIN_TRUNK=1` causes multi-minute CPU stalls under co-tenancy. `HALOGEN_FLASH_PIN_TRUNK=0` as last resort.
|
||||
- **Boot**: First request may be silent ~45-60s (first-touch KV pool).
|
||||
|
||||
### Llama.cpp (dankclaw/dankserver)
|
||||
- **Forks**: `ik_llama.cpp` (CUDA+Vulkan), `rocmfpx` (ROCmFP4).
|
||||
- **Flags**: `--cache-ram 16384` (medium), `--expert-precise-memory` (primary), `--max-model-len 200704`.
|
||||
- **MTP**: Qwen3.8-Flash-Next MTP head needs full per-layer loading (Flash-Next specific).
|
||||
|
||||
### VLLM (dankserver)
|
||||
- **Fork**: tcclaviger RDNA4 vLLM.
|
||||
- **Seqs**: `SEQS=3` mandatory (SEQS=2 fails `torch.fx` ConstraintViolationError).
|
||||
- **MTP**: Speculative decoding disabled/optional.
|
||||
|
||||
## 4. Operational Hygiene
|
||||
|
||||
### Bench Protocol
|
||||
1. `sudo systemctl stop <service>`
|
||||
2. Run benchmark
|
||||
3. `sudo systemctl start <service>`
|
||||
4. `curl -s <port>/health` → 200
|
||||
5. **Trap**: `trap EXIT 'systemctl start <service>'; curl -s <port>/health` for self-restoring harnesses.
|
||||
|
||||
### Quadlet Persistence
|
||||
- **Rootless Podman**: `restart: unless-stopped` only survives crashes. Bare containers vanish on reboot.
|
||||
- **Pattern**: Git-tracked `.container` files in `~/.config/containers/systemd/`.
|
||||
- **Historical**: Caddy lost after reboot (Aug 26); 5 observability services lost (Sep 11).
|
||||
|
||||
### Git & Deploy
|
||||
- **Repo**: `/home/blank/Projects/serverAppConfig`.
|
||||
- **Gate**: `deploy.sh` rejects dirty trees (untracked/modified files).
|
||||
- **Pattern**: `git stash` → deploy → `git stash pop`.
|
||||
|
||||
## 5. Troubleshooting Knowledge Base
|
||||
|
||||
### max_tokens 400 Overflow
|
||||
- **Symptom**: `prompt is X tokens and the context is Y, leaving room for Z`. `X+max > Y`.
|
||||
- **Fix**: Adjust `maxTokens` in `models.yml` (e.g., 57344) or `HALOGEN_CTX` (262144).
|
||||
- **Note**: OMP snapshots at process start; existing sessions need restart (`omp -r`).
|
||||
|
||||
### Orphaned `llama-server` (frank-medium)
|
||||
- **Symptom**: Unit stopped, GPUs busy.
|
||||
- **Diag**: `ps -eo pid,ppid,args | grep llama`, `rocm-smi --showmeminfo`.
|
||||
- **Fix**: `kill <pid>`, `systemctl start`, verify geometry.
|
||||
|
||||
### Context Bloat
|
||||
- **Source**: 175 managed-skills descriptions (~38k nonMessageTokens).
|
||||
- **Fix**: Distill to vault reference pages, move dirs to `~/.AGENTTRASH/`, purge on approval.
|
||||
|
||||
### Authentik User Creation
|
||||
- **Method**: `ak shell` on `dankclaw`, Django ORM (`User.objects.create`...), `ak shell` sets password, assign groups (`Non-Admin Users`, `jellyfin_users`).
|
||||
- **ACLs**: Group membership is sole gate.
|
||||
|
||||
### ComfyUI (iGPU)
|
||||
- **Target**: Strix Halo iGPU (3090 occupied).
|
||||
- **Quant**: FP8 or Q6.
|
||||
- **Display**: Ghostty `\e_G` base64 PNG, or `chafa` helper.
|
||||
|
||||
## 6. Skills & Playbooks
|
||||
|
||||
- **llm-fleet-model-swap**: Check HF model card, download sha256, commit → push → deploy.sh → health.
|
||||
- **halogen-flash-server-memory**: Memory model, two failure modes (stall, 400), co-tenancy.
|
||||
- **frank-primary-ram-stability**: DDR5 4800 vs 4600 MT/s, chunk boundaries, prefix cache eviction.
|
||||
- **frank-medium-bench-protocol**: Restore half optional? No. Self-restoring harness recommended.
|
||||
- **authentik-user-creation**: Django ORM workflow, group-based access.
|
||||
- **orchestration-quadlet**: Rootless Podman boot persistence.
|
||||
- **context-bloat-distillation**: Skill directory to vault migration.
|
||||
- **comfyui-qwen-workflow**: iGPU image generation, ghostty display.
|
||||
- **expert-arena-definition**: `arena = budget - non-expert - kv_reserve - graphs - overhead`.
|
||||
- **skill-routing-pitfall**: `html-change-report` for git commits, not creative writing.
|
||||
@@ -0,0 +1,100 @@
|
||||
- serverAppConfigs secret-scan baseline (2026-10-09, gitleaks 8.24 full history + manual sweeps, 559 files/1167 commits): tree and history CLEAN of real secrets. Known gitleaks false positives to expect on re-runs — (1) Bearer PASTE-YOUR-KEY in shared/homepage/docs.html (placeholder in public API guide); (2) curl -u admin:admin in serverAppConfigs-docs/Operations/Fleet Audit and Posture.md (Grafana default-credential RESIDUAL CHECK, not a leak); (3) FRANK_LARGE_*_SHA256 checksums in scripts/frank-*.env; (4) OAUTH_CLIENT_ID / GF_AUTH_GENERIC_OAUTH_CLIENT_ID in open-webui.container, grafana.container, dankclaw/compose.yaml — OAuth2 client IDs are public identifiers; the client SECRETS live only in git-ignored host .env files. Real-value secret file on disk: tests/e2e/.env.local (gitignored, verified). PENDING POSTURE ITEM reported to Frank, not yet approved for action: Grafana on dankclaw answers admin:admin on loopback 127.0.0.1:3002 (probe: ssh dankclaw curl -u admin:admin http://127.0.0.1:3002/api/org → 200) — rotation recommended, awaiting his go.
|
||||
- serverAppConfigs authentik user provisioning (five accounts kirby/ry/nic/debo/freya pk 17-21 created 2026-10-09; law now in vault commit 65dec6a): (1) After ak shell provisioning with the human-user template (groups Non-Admin Users + jellyfin_users + openwebui_users), OpenWebUI mints each account on FIRST SSO login with role pending (ui.default_user_role=pending in webui.db config table) — they log in fine, then cannot chat / picker is empty until UPDATE "user" SET role='user' WHERE email=…; the row does not exist before first login, so the flip is a first-login-sighting follow-up, not a provisioning step. No per-user model grant needed — the picker model carries the global access_grant row (model, frank-secondary, user, *, read). Written up in serverAppConfigs-docs/Dankclaw/OpenWebUI.md (SSO section). (2) The proxysession claim purge is only due when a group change follows an existing login; users created with their groups already in place have zero rows — verify with a per-user census instead of purging blindly. (3) The authentik-postgres container's DB role/db are authentik/authentik (POSTGRES_USER env), NOT postgres: podman exec authentik-postgres psql -U authentik -d authentik -tAc "select session_data-'claims'-'preferred_username', count(*) from authentik_providers_proxy_proxysession group by 1". (4) User.app_entitlements() requires the Application argument (Application.objects.get(slug="open-webui")) — bare call raises TypeError. Verification bar for a provision: authenticate(request=None, …) accepts the password and rejects a wrong one (the exact PasswordStage path) plus the open-webui entitlement resolves. _(context: serverAppConfigs 2026-10-09: provisioned five authentik human accounts on Dankclaw; vault pages Infrastructure/Authentik.md, Dankclaw/OpenWebUI.md, log.md updated (commit 65dec6a), bats suite 366 green.)_
|
||||
- radiance arch-capability audit (dankserver, pin 1.0.13, verified 2026-10-08): the container ships 6 arch plugins in /opt/radiance/share/radiance/architectures/ — qwen4exp_fp8, qwen35moe_fp8, qwen35_fp8, qwen35_bf16, llama_fp8, llama_dense_fp8. Capability check per plugin: grep -ac RADIANCE_CALIB / grep -ac gram / grep -ac mtp / grep -ac ngram on the .so (qwen35moe_fp8: calib=2 grams=2 mtp=4 ngram=0; gptq=0 in ALL plugins — GPTQ lives in rad-convert, not the plugin). Introspection pitfalls: the digest-pinned name (stilldeadcode/radiance@sha256:4c3f48…) does NOT resolve locally (podman tries a registry pull and fails); use the local tag docker.io/stilldeadcode/radiance:latest — image lives in ROOTLESS storage as dank (frank-primary runs rootless under the root unit; sudo podman ps is empty). Image has no ls/strings; podman exec frank-primary ls fails; correct form: podman run --rm --entrypoint sh docker.io/stilldeadcode/radiance:latest -c '…'. Verdict recorded: Qwen3.6-35B-A3B (qwen3_5_moe_text, 40 layers, 256 experts top-8, MTP×1) is plugin-supported; w4dn8 recipe maps minus PLE/HC terms; ~20 GiB w4+i8 = fully VRAM-resident on ONE 32 GiB R9700 (no TP=2/host-pool/disk-tier/MNBT-2048 law); convert ~1–1.5 h. Full audit in serverAppConfigs-docs/LLM/Quantization Production.md. _(context: serverAppConfigs 2026-10-08: feasibility audit of porting the frank-archon w4dn8 radiance recipe to Qwen3.6-35B-A3B; KB page Quantization Production.md carries the full write-up.)_
|
||||
- Quantization production research (2026-10-08, full detail in repo page serverAppConfigs-docs/LLM/Quantization Production.md, linked from both vault indexes): (1) Law — production is hardware-agnostic (imatrix/GPTQ/ablation/pruning are torch-CPU work); the card only picks what it can DECODE. (2) Frank's framing: the 3090 is the ONLY unlocked fleet device — llama.cpp there is a choice; sm_86 decodes GGUF K/I, GPTQ/AWQ-Marlin INT4, SmoothQuant INT8, weight-only FP8 (FP8 W8A8 compute is Ada/Hopper SM≥89 only; NVFP4 Blackwell; 3090 has NO FP8/FP4 silicon), EXL3 (ExLlamaV3, convert.py runs on-card). The 3090 is the natural quant R&D bench (produce on host, decode+score vs fleet KLD gates). (3) REAP (Cerebras, arXiv 2510.13999, github CerebrasResearch/reap) expert saliency = router gate-values × activation energy — the INVERSE of our derive-protected-experts.py energy-share law; REAPing Flash-Next experts 25–50% re-enters the radiance pipeline at calibration, zero new tooling. (4) HauhauCS K_P "Perfect" = imatrix on abliterated weights + per-model per-tensor mixed-type profile via llama-quantize --tensor-type (plain GGUF, no fork; Q8_K_P census: Q8_K+F32+F16); reproducible with binaries built by scripts/build-llama-rdna4.sh. (5) Heretic (p-e-w/heretic, AGPLv3) = automated abliteration (refusal-direction ablation + layer dropping, refusals-vs-KL tuned), NOT a quantizer; abliteration (Arditi 2406.11717) is rank-1 fp-editing. Order law: surgery → calibrate → quantize; keep refusal tokens out of calibration. (6) .hgn production door is CLOSED — no public quantizer exists in peonist-ai repos; self-quant on Strix runs through GGUF runtimes. _(context: serverAppConfigs research session 2026-10-08; scouts + first-hand verification of REAP/Heretic/abliteration/HauhauCS K_P and per-runtime quant matrices; KB page Quantization Production.md committed to main.)_
|
||||
- serverAppConfigs: (1) The quantization-production research (2026-10-08) lives in serverAppConfigs-docs/LLM/Quantization Production.md — per-card playbooks (3090 imatrix/llama-quantize; R9700 radiance rad-convert pipeline is fully self-owned; Strix .hgn production is CLOSED, no public quantizer) and the verified prior-art families: REAP (Cerebras expert pruning — saliency = our derive-protected-experts.py energy-share statistic, inverse polarity; REAPing Flash-Next's 49,152 experts 25-50% drops the w4 resident set to ~58-87 GiB and kills the disk tier), Heretic (p-e-w/heretic = automated abliteration + layer dropping, NOT a quantizer), abliteration order law ablate→calibrate→quantize, HauhauCS K_P = plain GGUF with imatrix + per-tensor mixed-type profile (llama-quantize --tensor-type), no fork needed. (2) Docs seam law: tests/docs_completeness.bats requires EVERY page under serverAppConfigs-docs/ to be linked from the ROOT index.md (serverAppConfigs-docs/index.md), not just the section index (LLM/index.md) — a new vault page linked only in LLM/index.md fails seam 2. _(context: Quantization Production research turn, 2026-10-08; suite caught the root-index link requirement)_
|
||||
- Dankclaw homepage (serverAppConfig shared/homepage) durable facts from 2026-10-08 wave 8: (1) Smoke-server trap escalation: when a localhost http.server smoke check serves stale content, the authoritative diagnostic is for p in $(pgrep -f "http.server 8099"); do readlink /proc/$p/cwd; done — a listener from an OLD staged tree can survive and serve while your new files sit untouched elsewhere; a ss | grep || (start) guard silently skips the restart when the stale listener is found. Kill the exact pid, restart from the fresh tree, re-curl for a new-file marker. (2) Warning UI convention (Frank-approved): all privacy/uncensored warnings are native details disclosure chips ("⚠ privacy note ▸/▾"), collapsed by default, pinned to the box bottom via margin-top:auto in a flex-column panel/card; the clickable-panel open handler MUST return on ev.target.closest('a, .svc-warn') or the chip tap navigates to the app. (3) The docs page is served at path /docs/ — a bare /docs.html URL hits the SPA fallback and renders the homepage shell. (4) frank-archon runs the fleet's OWN payload: homepage card + docs Model bullet link https://huggingface.co/frank8465/Swift-1.5-Qwen3.8-Flash-Next-GPTQ-W4A8-Radiance-AMD-R9700 (published on HF, verified via API); spec = in-house GPTQ-w4 quant of Swift1.5-Qwen3.8-Flash-Next, ~114 GiB, three submissions (--max-num-seqs 3). The other two cards' HF links are their publishers' upstreams (accurate — verbatim artifacts). (5) OpenWebUI 0.6 API-key flow for docs/screenshots: avatar (bottom-left) → Settings → Account → API keys → Show → copy icon (⋯ rotates); no "Create new secret key" button exists. _(context: serverAppConfig homepage wave 8: bottom-pinned disclosure chips, live OWU api-keys screenshot, archon HF publication link; smoke verification against staged trees on the dev box.)_
|
||||
- HF publication completed 2026-10-08: (1) Model repo frank8465/Swift-1.5-Qwen3.8-Flash-Next-GPTQ-W4A8-Radiance-AMD-R9700 — public, 15 files, .rad = qwen4exp-swift15-w4dn8-i8-hc8m.rad 122,071,973,888 B sha256 09dd9306e859… uploaded via hf_xet at ~46 MB/s in ~29 min from dankserver (systemd-run --user --collect transient unit + .rc sentinel). (2) Dataset frank8465/swift15-gptq-w4-calib-corpus — public, CC BY-SA 4.0, anonymized calibration corpus (calib.jsonl 1,294 prompts, kld.jsonl 197, manifest, make-corpus.py, scrub-corpus.py); original unpublished hashes calib edb9b745…, kld 6e7df2f5…; cross-linked with model card (§2/§4 links, 4 refs). Scrub law: fleet slices are agent-session transcripts — scrub substring-class (not word-boundary): dankserver/dankclaw/frankbook/franksdank (bare brand = container image namespace franksdank/*), franksdank\.(cloud|casa), /home/dank, emails; corpus carries NO real credentials (92 FRANK_SUDO_PW hits = variable refs; private-key hits = bats template headers; pw-assign hits = upstream test fixtures). HF card validator law: custom licenses must be license: other + license_name + license_link (raw SPDX-unknown value = ValueError at upload_file, AFTER sha gates). hub 1.28 API: upload_file/create_commit kwargs keyword-only; HfApi has no rename_file/hf_hub_url/commit_api attrs; whoami hides canWrite for legacy tokens (probe = create_repo exist_ok=True). Secret storage pattern: ~/.config/secrets/name 600 + .zshrc loader export, token mirrored to ~/.cache/huggingface/token; write token installed on dev box + dankserver.
|
||||
- Dankclaw homepage/docs (serverAppConfigs, deployed 2026-10-08): (1) Frank's warning doctrine — hover tooltips are DEAD (unreachable on touch); warnings render as native details disclosure chips ("⚠ privacy note ▸") collapsed by default, pinned to the box bottom via margin-top:auto in flex-column .panel/.mi-card; the clickable-panel open handler must ignore .svc-warn taps (chip taps bubble → navigation bug, fixed). Same chips on docs.html (privacy + uncensored-model). (2) Live-verification trap: https://www.franksdank.cloud/docs.html is swallowed by the SPA fallback and serves the homepage shell — the docs page lives at /docs/. (3) Served model ids on the API layer are STILL frank-primary/frank-large/frank-secondary (/v1/models live-verified) — the frank-archon/dragoon/zealot rename is harness+homepage+docs only; a served-model-name flip (needs inference restart + OpenWebUI picker ACL re-arm) is offered to Frank, not yet approved. (4) OWU 0.6 API-keys screenshot flow (headless, SSO profile, QA E2E Bot account): click img[src*="profile/image"] → menu Settings → tab Account → button Show → scroll the settings dialog (largest scrollHeight element) to bottom; the screenshot({path}) option did not write a file — grab the returned omp-sshots webp artifact and ffmpeg-convert to PNG. _(context: serverAppConfigs homepage wave-8b: disclosure chips, docs page mobile overhaul, live OWU screenshot, 2026-10-08)_
|
||||
- HF corpus publication staged 2026-10-08 (NOT yet published — needs Frank's explicit go): dataset repo id frank8465/swift15-gptq-w4-calib-corpus, uploader /home/dank/hf-swift15-corpus/upload-corpus.py (dry-run default, --go arms; radcorpus venv), package = README.md datasheet (CC BY-SA 4.0 for the wikitext share-alike; tiny-codes permissive; fleet slice ours), manifest.json (paths replaced with source dataset names), anonymized calib.jsonl (39,236,524 B, sha256 3e79837f…) + kld.jsonl (5,702,825 B, sha256 b5b70ae9…), scripts/make-corpus.py (literals anonymized) + scripts/scrub-corpus.py. The scrub is deterministic (13 rules: LAN→10.x, dankserver/dankclaw/frankbook→server-a/server-b/client-c, franksdank.cloud|.casa + bare brand→example, /home/dank→/home/user, emails→user@example.com, FRANK_SUDO_PW→DEPLOY_SUDO_PW); audited-clean findings: corpus carries NO real credentials (92 FRANK_SUDO_PW hits = variable refs; private-key hits = bats template headers; password hits = upstream fixtures), the September HF-token leak did NOT reach the corpus; residual probes after scrub are self-replacement/encyclopedic-content noise. Originals stay private, receipted in the datasheet: calib sha256 edb9b7456dbeb8d7def21380778988201aead1cf3f073d3266aaa1447c437603, kld 6e7df2f562d712e04a669220a0c76788688aca5548867596b4e070902937b750 — published corpus is method-reproducible, deliberately NOT byte-identical to the shipped .rad's calibration input. Model repo frank8465/Swift-1.5-Qwen3.8-Flash-Next-GPTQ-W4A8-Radiance-AMD-R9700 was created public and its 13 small files uploaded; the 113.7 GiB .rad streams via hf-upload.service (log /home/dank/logs/hf-upload.log, rc sentinel /home/dank/logs/hf-upload.rc). Pending after stream: repo census, dataset --go, model-card §4 dataset link, runbook/AGENTS link. Homepage commit c4dca68 (AI guide link) undeployed — Frank's uncommitted docs.html + openwebui-api-keys.png keep the tree dirty and deploy.sh refuses dirty trees. _(context: serverAppConfigs 2026-10-08: model+corpus publication campaign state at end of session; dataset upload awaits user go.)_
|
||||
- Dank fleet secret layout (2026-10-08 cutover): all shell secrets live in ~/.config/secrets/ (dir 700, files 600) on BOTH the dev box (blank) and dankserver (dank) — kagi_api_key (dev), hf_token (both; the write-scope fine-grained token is the canonical value, Frank's copy named hf_token_write on dev). The .zshrc files carry ONLY loaders of the form [ -f ~/.config/secrets/X ] && export X="$(cat ~/.config/secrets/X)" — NEVER literal secrets in rc files (the pre-cutover literals on dev .zshrc:3 Kagi and dankserver .zshrc:40 HF were extracted; dankserver keeps .zshrc.bak-20261008). ~/.cache/huggingface/token (600) is kept in sync on dankserver — it is what huggingface_hub and the upload scripts read. Drop-in procedure for a new HF token, on dankserver: read -rs KEY; printf '%s' "$KEY" | tee ~/.config/secrets/hf_token ~/.cache/huggingface/token; chmod 600 both files — value never in chat/argv/history. HF card-metadata law: license: must be a recognized SPDX id — custom licenses need the trio license: other + license_name: + license_link: (the swift-open-license-1.0 upload died on this; the ukisai base card uses the trio). _(context: serverAppConfigs 2026-10-08: HF publication campaign; token-scope 403 led to rc-file secret audit and remediation on both hosts.)_
|
||||
- Frank's secret-storage convention (established 2026-10-08, both dev box + dankserver): secrets live in ~/.config/secrets/ (dir 700, files 600; HF write token as hf_token_write, canonical copy as hf_token synced to ~/.cache/huggingface/token); .zshrc carries ONLY loaders like export HF_TOKEN="$(cat ~/.config/secrets/hf_token 2/dev/null)" — never literal secrets in rc files (a literal HF token sat in dankserver .zshrc:40 since 2026-08-18 and a literal Kagi key in dev .zshrc:3; both defused this day, .zshrc.bak-20261008 kept). New tokens are dropped in by Frank himself with read -rs — value never touches chat, argv, or history. Hub API facts: huggingface_hub 1.28 whoami is uninformative for fine-grained tokens (type access_token, canWrite/scopes absent) — the decisive write-scope probe is create_repo(real-target, exist_ok=True) (403 = no write; success proves scope and pre-creates); create_repo returns bare RepoUrl with no .private attr. Corpus audit (swift15 calib, /home/dank/corpora/out): calib.jsonl 1,294 prompts 39.5 MB = tiny-codes 45% (permissive/MIT) + wikitext-103 30% (CC BY-SA 4.0 share-alike on derived data, attribution; does not reach model weights) + fleet 25%; fleet+KLD slices are agent session transcripts with pervasive infra fingerprints (47,969 LAN-IP hits, 663 franksdank, 13,644 /home/dank; zero credential VALUES incl. FRANK_SUDO_PW) — scrub vs byte-faithful is Frank's pending ruling; recommended route: scrubbed corpus + scrub script + original's sha256 receipt.
|
||||
- Dank fleet secret convention (est. 2026-10-08): literal API keys/tokens never live in rc files. Canonical storage = ~/.config/secrets/name (dir 700, file 600) on each host; .zshrc lines are loaders (export KEY="$(cat ~/.config/secrets/key)"). Fixed on 2026-10-08: dev box .zshrc:3 KAGI_API_KEY literal and dankserver .zshrc:40 HF_TOKEN literal (agent-added 2026-08-18) extracted to secret files + loaders; dankserver kept ~/.zshrc.bak-20261008. HF token canonical file ~/.cache/huggingface/token (600) is kept in sync from the secret file and is what the upload scripts read. HF token state 2026-10-08: frank8465's cached token is a legacy read-scoped access_token — repo create fails with 403 "don't have the rights to create a model under the namespace"; Frank is minting a fine-grained WRITE token (needs repos:write + datasets:write); drop-in = read -rs KEY; printf '%s' "$KEY" | tee ~/.config/secrets/hf_token ~/.cache/huggingface/token; chmod 600 … run by Frank himself on dankserver (token must never enter chat/argv/history). Payload corpus audit 2026-10-08: calib/kld corpora contain ZERO literal secrets (all 92 FRANK_SUDO_PW hits are variable references); fingerprints present are /home/dank paths, dankclaw/dankserver names, LAN IPs (publishing the corpus needs a scrub pass). HF model-card payload (README/scripts/artifacts) verified clean: no credentials, no IPs/hostnames.
|
||||
- Swift1.5 GPTQ-w4 campaign corpus (dankserver /home/dank/corpora/out/, manifest.json seed 20261006): calib.jsonl = 1,294 prompts, mix 45 % code from nampdn-ai/tiny-codes (permissive, attribution) / 30 % prose from Salesforce/wikitext-103 (**CC BY-SA 4.0 — any published derivative dataset must be CC BY-SA + attribution**) / 25 % tools harvested from serverAppConfigs (own content); kld.jsonl = 197 prompts from local repo material only, pools disjoint from calib. Frank approved publishing the corpus as an HF dataset (45 MB; verdict: closes the reproducibility loop StillDeadcode left open — their card ships the recipe with $CALIB unpublished). Gates before publishing: scrub fleet infra fingerprints from harvested slices; license manifest per slice; fine-grained token must carry dataset-write. Token law (measured 2026-10-08): the cached dankserver token is READ-scoped — hf whoami succeeds but create_repo 403s "no rights to create under frank8465"; an unauthenticated API 404 on a repo path is indistinguishable from private (HF returns 401 "Invalid username or password" for both); authed 404 = repo truly absent. Frank must mint the write token himself (write via read -rs on dankserver, never chat/argv). HF upload campaign staging: frank8465/Swift-1.5-Qwen3.8-Flash-Next-GPTQ-W4A8-Radiance-AMD-R9700, payload 122,071,973,888 B sha256 09dd9306…, staging /home/dank/hf-swift15-upload, uploader upload-to-hf.py (dry-run default, --go), launch as transient unit hf-upload.service.
|
||||
- On the Dank fleet hosts (systemd-logind, KillMode=control-group), setsid nohup CMD & launched from ssh host '…' is SIGKILLed silently ~seconds after the ssh command returns — logind deactivates session-N.scope and kills the whole cgroup; setsid/nohup do not escape the cgroup, and there is NO journal/OOM trace of the death. The reliable detached-launch pattern is ssh host 'export XDG_RUNTIME_DIR=/run/user/$(id -u); systemd-run --user --collect --unit=NAME -p SendSIGKILL=no -p StandardOutput=append:/path.log -p StandardError=append:/path.log sh -c "CMD; echo \$? /path.rc"' — a transient user unit outside the session scope (works because Linger=yes), with the rc sentinel written by the unit itself. Diagnose by: ps -p PID (gone), log file 0 bytes (buffered stdout lost on SIGKILL; python needs -u), and journalctl --since … | grep "session-.*Deactivated" timestamps matching the death window. Hit 2026-10-08 launching the HF 113.7 GiB upload from dankserver.
|
||||
- Radiance Swift1.5 quant baselines (verified 2026-10-08 from StillDeadcode/swift1.5-qwen3.8-next-flash-fp8-iq4r-moe card + HF API): engine author's file = 122,013,646,848 B; recipe core identical to frank's (gptq table=w4nl group=64 scale=fp8_e4m3 scale2_value=2^-13 transform=fwht128 cd=3; int8 trunk; fp8 HC; PLE fixed scale 1.99317932128906e-4) — differences: 1M vs 10M calibration, 16 protected experts total vs 24/layer (1,152). THEIR FILE WAS NEVER MEASURED against own bf16; card quotes base-model recipe numbers: mean KL 0.0789 all-tokens / 0.0239 text-only, top-1 92.85/94.57 %, and the two-bf16-implementations reference floor is 0.0399. Frank's measured file: KL 0.01314 / top-1 97.92 % — beats the bf16-implementation noise floor ~3x; 97.9 % top-1 ≈ Q6_K-band agreement on the GGUF ladder (no public GGUF ladder exists for this arch, KL is the honest yardstick). Use these as standing comparison baselines for future Swift1.5 quant campaigns. Model-card doctrine Frank approved for HF publications: hard-first ordering (technical content leads, a bold "In plain English:" paragraph CLOSES each section); Randall-Munroe voice (dry concrete humor, no marketing, no AI-isms); include a "Relationship to the engine author's release" section with the measured table; state receipts as exact bytes. _(context: HF publication of frank8465/Swift-1.5-Qwen3.8-Flash-Next-GPTQ-W4A8-Radiance-AMD-R9700 (upload launched 2026-10-08); comparison research against the radiance author's own quant of the same checkpoint.)_
|
||||
- Comparison facts vs StillDeadcode/swift1.5-qwen3.8-next-flash-fp8-iq4r-moe (the radiance author's own quant of our base checkpoint ukisai/Swift1.5, fetched 2026-10-08 via HF API): their .rad = 122,013,646,848 B (ours 122,071,973,888 → +55.6 MiB / +0.048 %). Expert GPTQ rule is byte-identical to our staged recipe (gptq table=w4nl group=64 scale=fp8_e4m3 scale2_value=0.0001220703125 transform=fwht128 rule=search cd=3), as are int8 trunk group-128, fp8 e4m3 HC, PLE fp8 fixed scale 1.99317932128906e-4, and the inert u2 mtp.draft_head rule. Differences: their calibration 1M tokens (ours 10M); their protected set = 16 experts TOTAL (base-model rule transplanted; card example layer4/expert108 = 15.8 % energy from 0.18 % routing) vs our 24/layer = 1,152 (energy ≥3 % AND ≥3× routing, parity-padded). Their card states the file "was not measured against its own bf16 weights" — quoted KL 0.0789 all-tokens / 0.0239 text-only, top-1 92.85/94.57 % are the recipe-on-base numbers, with their two-bf16-implementations reference row at 0.0399 mean KL. Ours measured on-file: 0.01314 / 97.92 % → ~3× better than their bf16-implementation noise floor. Frank's framing (adopted in card §2 "Relationship to the engine author's release"): 93–94 % top-1 ≈ Q4-class, 97.9 % ≈ Q6_K band; "the recipe is theirs, the extra numbers are ours."
|
||||
- Frank's homepage doctrine updates (2026-10-08, eighth wave — supersedes the octagon-cut corner rule): (1) NO clip-path chamfers on bordered boxes — clip-path slices the 1px border and the corners "feel like gaps"; use border-radius (panels/mi-card 6px, model-info 8px, badge 3px, buttons/callouts 4px). (2) Section-title diamond must be a fixed-width inline glyph (inline-block, width 18px, margin-left -24px to eat the 24px padding), never absolutely positioned (drifts/wraps). (3) No hover-only content: privacy notes (.svc-warn/.mi-warn) render FULL text inline — title-attribute tooltips are unreachable on touch; the "hover to read" pattern is retired. (4) Closed WIP items grey out via :has(.badge.wip-closed) — dim #5d87a8 name/desc, muted border, art grayscale(0.75) brightness(0.7); WIP: DOWN keeps the red fault treatment (down=fault, closed=status). (5) Homepage data facts: services.json "wip" is a STRING ("down"/"closed"), not boolean — wip badges come from services.json, status.json only drives non-wip probes; the Dragoon tagline is ONE spliced file sfx/dragoon-tagline.ogg (loudnorm -16 LUFS/TP -1.5 per track, silence-trimmed, 350 ms beat, limiter; the two SC2 source oggs trashed). Voice playback proves via tab.requests() 206 + mi-speaking class (detached new Audio, never in DOM).
|
||||
- Frank's doctrine for public-facing docs (model cards, READMEs): every section must open with a bold-lead "**In plain English:**" paragraph aimed at a non-technical-but-educated reader, and the overall voice must read like Randall Munroe — short declaratives, concrete analogies (cartridge vs DVD, laminating the pages you touch daily), jokes only in the margins, no AI cadence (no delve/seamless/moreover, no bullet-spam, no bold-everything). Technical facts, numbers, and verbatim receipts stay untouched under the plain layer. Applied to the HF card staged at ~/Projects/hf-swift15-upload/README.md (final, synced to dankserver /home/dank/hf-swift15-upload/, gates green: sha256 MATCH on the 122,071,973,888 B .rad = 09dd9306e859f5343d9cb77ee6736636ab9df50f3ed2153350dfbb35be5e6948; publication awaits Frank's --go of upload-to-hf.py). Card prose deliberately omits the w4dn8 relic lore and the mnbt-3072 gate; staged scripts keep their verbatim fleet names as provenance receipts — renaming staged copies would falsify the chain.
|
||||
- HF publication-card laws (frank8465 Swift1.5 w4 campaign, 2026-10-08): (1) The published .rad filename is a creation-time choice (Swift-1.5-Qwen3.8-Flash-Next-GPTQ-W4A8-Radiance-AMD-R9700.rad) — no re-convert needed to rename for readers; the old internal name survives only in the .rad's baked storage.filename metadata (engine ignores it; visible in rad-info artifacts) and in staged toolchain scripts, which stay verbatim fleet receipts — renaming staged copies would forge provenance. Final card carries ZERO relic-name prose except the convert script filename; Frank's doctrine: reader-facing files get clean names, provenance receipts stay untouched. (2) Card style preference (Frank, final review): drop engine-shape bookkeeping readers won't care about (dual-gate MNBT comparison rows), EXPAND the interesting engineering (protected-bf16-expert law: energy-share ≥3% AND energy/routing ≥3×, even parity padding per rank half, lands 24/512 per layer = 1,152 experts = 2.72 GiB). (3) Preview trap: Frank screenshots Chrome with DevTools device-emulation ON — the "fullPage" badge clips tall pages at the emulated viewport height and looks like truncated content; check the served HTML tail first, then say Ctrl+Shift+M. (4) Line-numbered PUT edits on a script shifted by earlier inserts silently CLOBBER the line at the old number (killed two module constants in upload-to-hf.py; caught by dry-run, NameError after the sha gate) — re-view the region in the returned anchored view after every multi-edit sequence, and grep-verify constants before shipping. _(context: Swift1.5 GPTQ-w4 HF publication staging, 2026-10-08: card finalized, dry-run gates green, awaiting Frank's --go)_
|
||||
- dankserver deployed payload receipt: /home/dank/models/rad/qwen4exp-swift15-w4dn8-i8-hc8m.rad is 122,071,973,888 B with sha256 09dd9306e859f5343d9cb77ee6736636ab9df50f3ed2153350dfbb35be5e6948 (computed 2026-10-08 with sha256sum on the host, verified twice by the upload script's gate). scripts/frank-primary.env's FRANK_PRIMARY_SHA256 pins only the fp8-native rollback file, not this w4dn8 one — this is the only recorded receipt for the active payload. HF publication staged as frank8465/Swift-1.5-Qwen3.8-Flash-Next-GPTQ-W4A8-Radiance-AMD-R9700 via /home/dank/hf-swift15-upload/upload-to-hf.py (dry-run default, --go arms it). _(context: serverAppConfigs 2026-10-08: HF publication staging for the deployed frank-primary w4dn8 payload; upload pending Frank's --go. Staging at ~/Projects/hf-swift15-upload (dev) + /home/dank/hf-swift15-upload (dankserver).)_
|
||||
- HF Hub facts for frank8465 (free plan), verified 2026-10-08 via whoami API + hub docs: (1) Free private repos are hard-capped at 100 GB total file storage (all private repos combined); exceeding it disables private repos at renewal, public repos unaffected. (2) Public repo storage on the free tier is best-effort with NO published limit — precedent: a public model repo with a 121.97 GiB single file lives on the free tier. (3) hf auth whoami --output-format json on dankserver's token shows name/fullname/orgs[]/plan (plan None = free) — and access tokens CANNOT list/write datasets (write op → 403 "Invalid credentials in Authorization header"; read → 401), so create_repo must use repo_type="model". (4) NEVER pass a Bearer token in a curl command line — argv is world-readable in /proc/pid/cmdline (ps/pgrep leak it); use curl -K mode-600 file with header = "Authorization: Bearer …". (5) Gated bases (ukisai/Swift1.5) are gated at the CONVERSATIONAL layer only: base weights, tokenizer, AND the MTP tensors ship openly; the ablation variant Quat3rnion/halogen-…-v2-abliterated is ungated. (6) The fleet radiance w4 payload: /home/dank/models/rad/qwen4exp-swift15-w4dn8-i8-hc8m.rad, 122,071,973,888 B, sha256 09dd9306e859f5343d9cb77ee6736636ab9df50f3ed2153350dfbb35be5e6948, KLD mean 0.01314 / top-1 97.92 % / ppl ratio 1.0052 vs BF16. Upload staging: /home/dank/hf-swift15-upload/ + upload-to-hf.py (dry-run default, --go arms, radcorpus venv with hf_xet). _(context: Swift1.5 GPTQ-w4 .rad publication campaign on frank8465, October 2026)_
|
||||
- radiance container introspection: the image entrypoint is radiance, so podman run ... image /opt/radiance/bin/rad-info FILE fails with "E unknown option" — the path is parsed as a radiance flag. Correct form: podman run --rm --entrypoint /opt/radiance/bin/rad-info -v /home/dank/models:/models:ro image /models/rad/FILE.rad. rad-info prints: RAD2 header (bytes, tensor/plane counts), a 181-key metadata block (n_layers, n_expert, full_attention_interval, n_ctx_train...), and an encoding census with per-encoding tensor counts, sizes, and % share — the authoritative receipt for a .rad (gptq params incl. calib path, group, table, transform; rtn scale values; cast bf16 lines). No strings(1) in the image; dump to a temp file and grep it — zsh over ssh mangles sed -n "X,Yp" with $( ) substitutions. _(context: serverAppConfigs 2026-10-08: model stat sheet for frank-primary's qwen4exp-swift15-w4dn8-i8-hc8m.rad came from rad-info census (gptq w4 59.17 GiB/52.1%, PLE fp8 47.68 GiB/41.9%, i8 trunk 3.41 GiB, bf16 2.72 GiB).)_
|
||||
- radiance swift15 w4 cutover laws (Dankserver frank-primary, live 2026-10-08, main a30ac94): (1) w4 expert tables stream from the file tier IN PLACE only at --max-num-batched-tokens ≤ 2048; at MNBT 32768 the engine demands a widened form and the disk tier refuses ("DID NOT FIT ... stored form is not the container's") — a KLD/eval gate must run at an MNBT the w4 file can serve, and chunk size is softmax-invariant (two PASSes 0.0146 vs 0.0131 mean-KL at MNBT 3072/2048 bound the float-noise band at ~0.0015). (2) The radiance harness contract survives model swaps via --served-model-name "$FRANK_PRIMARY_SERVED_NAME" (=frank-primary): /v1/models id is independent of the .rad filename — models.yml needs no edit on payload swaps. (3) frank-primary prod shape at cutover: fp8 KV + spec depth 2 + seqs 3 + vram pair 26071/3543 (KLD gate scored at fp8 KV, gate and boot agree); decode ~190 t/s live (512-token streamed probe, 30.8 gate). The int8+spec3 campaign arm (136 t/s @44.5k) is NOT what shipped — verify env against the deployed container argv (podman inspect ... --format {{join .Config.Cmd " "}}, rootless as dank) before writing docs. (4) radiance /v1/completions SSE: guard payload=="[DONE]" before json.loads; there is no bare /completion (404). (5) deploy.sh ff-pull on dankserver fails on untracked campaign files (autoresearch-frank-primary.sh) that the merge wants to track — move to ~/.AGENTTRASH first. (6) kld-gate-swift15.sh edits nearly clobbered set -euo pipefail/PIN= by a comment PUT: never place body PUTs over script prologue lines. _(context: serverAppConfigs swift15-radiance-w4dn8 campaign completion: final convert → KLD gate (KL 0.0131, top-1 97.92%) → deploy → live verification, 2026-10-08.)_
|
||||
- Dank fleet (serverAppConfigs, 2026-10-08): frank-primary Swift1.5-GPTQ-w4 campaign gates ALL PASS — KLD mean KL 0.0131 / top-1 97.92 % (two shapes: MNBT 3072 → 0.0146, MNBT 2048 → 0.0131; the Δ≈0.0015 is the float-order noise band of chunk size, so KLD verdicts are chunk-size invariant), throughput 136 t/s median @44.5k vs 30.8 bar, fleet probes green, 363 bats green on merged main (02fa1e2+). Radiance laws added: (1) w4 expert tables stream from the file tier IN PLACE only at MNBT ≤ 2048 — at MNBT 32768 the engine demands a widened form and the disk tier refuses ("DID NOT FIT … stored form is not the container's"); any offline mode (KLD scoring, calibration) on a disk-tier w4 model must run MNBT ≤ 2048 with a big host pool (40960). (2) A KLD gate must mirror the reference RECORDING shape (tp, pool, MNBT, kv dtype, deterministic) — shape deltas ride on the KL as noise. Watcher lesson: a log-path watcher racing a log truncation of a relaunched job can match the PREVIOUS run's verdict — anchor patterns (^KLD GATE PASS$) and grep for verdict lines the final writer emits, not substrings the launcher echoes. _(context: swift15-radiance-w4dn8 cutover on dankserver; merge to main 02fa1e2, deploy pending at capture time)_
|
||||
- Radiance w4 disk-tier laws (measured on dankserver swift15 campaign, 2026-10-08): (1) GPTQ w4 expert tables stream from the file tier IN PLACE only at --max-num-batched-tokens = 2048; at larger MNBT the engine demands a widened form and the placement step dies with "DID NOT FIT ... stored form is not the container's (a relayout, a widening or a shard)" — production MNBT 2048 is a kernel law, not a tuning choice. (2) A KLD-scoring boot must mirror the reference RECORDING shape (same pin, tp, kv dtype, headroom, pool, MNBT) because chunk size changes float order that rides on the KL floor; the reference recording ran MNBT 32768 (bf16 model, no kernel constraint), so candidate scoring at 2048 accepts float-order noise as documented deviation. (3) host-pool sized for a bf16 model (12 GiB) under-serves the w4 candidate — the w4 gate_up slab needs the production pool (40960). (4) The calibration tap lives in /opt/radiance/share/radiance/architectures/qwen4exp_fp8.so, not the radiance binary — grep -ac the .so before declaring a flag missing (the image has no strings(1)). _(context: serverAppConfigs swift15-radiance-w4dn8 cutover: KLD gate boot failed twice (pool 12288 DID NOT FIT at MNBT 32768) before shape was matched to recording + w4 kernel law; scripts/kld-gate-swift15.sh encodes the final shape.)_
|
||||
- Frank's homepage aesthetic doctrine (2026-10-08, after three theme waves): StarCraft Protoss console = flat deep-navy (#050d16) panels with octagon-cut corners on crisp 1px steel borders, warp-blue/psychic-cyan accents used sparingly, Share Tech Mono body + Rajdhani 600 headings (Google Fonts, linked in all three HTML pages: index/admin/docs), colored app icons. His verdict on the maximalist wave: "too busy — clean up and simplify/solidify" — so: no scanlines, no lattice backdrop, no ambient text glows, no inner frame lines, no blinking cursors, no pulsing states; motion is reserved for state only (speaking-voice pulse). Default to minimal on this page; maximal effects get rejected. Docs-page rules from the same session: family docs order = getting-in → troubleshooting → media → AI sections last; the *arr stack is admin-only and never mentioned in family docs; each fleet model carries an HF source link (services.json model.hf + docs Model: bullets). Smoke-test pitfalls on this repo: (1) a python http.server left on port 8099 from an earlier wave serves the OLD tree — pkill -f "http.server 8099" before re-serving a fresh copy; (2) the shared headless browser disk-caches the pages — every verification URL needs a fresh ?v=timestamp query, and style.css/app.js are cached too (the bust propagates via the page URL). Deploy flow unchanged: tests/run.sh (the frank_primary_unit MAX_SEQS failure is Frank's campaign, hands-off) → deploy.sh dankclaw --service homepage → e2e gate.
|
||||
- Dankclaw homepage (serverAppConfig shared/homepage) aesthetic direction from Frank, 2026-10-08: StarCraft Protoss console — deep space navy #050d16 bg with warp-blue #3fa9ff/psychic-cyan #7fe7ff accents; amber CRT palette explicitly ABANDONED ("abandon amber"). Cornering: MAXIMAL — octagon-cut (all four corners chamfered 18–24px) on panels/cards/badges/buttons with a second inset octagonal frame line (::before double-bevel). Fonts: body "Share Tech Mono", headings "Rajdhani 600" (Google Fonts); VT323 retired ("more legible yet retro-clear terminal font… Think Starcraft"). Tiles must carry pictograms, NOT letter monograms — initials-on-hex icons read as "reverted" and get rejected (app-*.svg: book/GameBoy/folder/crumbling-wall). Two verification gotchas learned this session: (1) a reused python http.server smoke port serves browser disk-cache — always append ?v=epoch cache-bust when re-verifying, and restart the server from the fresh tree (a second python3 -m http.server on a busy port silently fails; the OLD tree keeps serving). (2) deploy.sh homepage runs the Playwright e2e gate ([3/3]); the known pre-existing failure frank_primary_unit.bats MAX_SEQS is Frank's radiance-campaign hands-off zone — never fix it. _(context: Waves 4–6 of the homepage theme work on 2026-10-08: Protoss theme, HF links, navy recolor; live census verified at commit 90f6785 with e2e 34/34.)_
|
||||
- Verifying static frontend changes via a scratch python3 -m http.server + headless browser on the Dank fleet has two traps that fake the result: (1) a second python3 -m http.server same-port launch silently fails when the port is occupied — the OLD server keeps serving the OLD tree copy, so the browser census reports stale state and a fresh copy looks broken (observed: fresh docs.html served 0 HF links while curl on the same URL proved the new file was live); (2) the headless browser disk-caches heuristic-cached responses (http.server sends no Cache-Control), so re-navigating to the same URL after re-syncing files returns the cached page. Protocol: kill any prior server on the port (pkill -f "http.server port") before starting the fresh one, verify the server actually serves the new tree with curl | grep -c new-marker, and cache-bust browser navigation with ?v=timestamp. Also: bash tool calls with async + name + timeout can fail with "Service mode does not accept async or timeout"; plain async works. _(context: 2026-10-08 serverAppConfig homepage waves: stale smoke5 server on :8099 + browser cache produced two false-negative verification rounds before curl proved the tree was fresh.)_
|
||||
- Dankclaw homepage client-side identity gating: Caddy's post-auth listener (http://:8017, homepage host block) can echo the Authentik outpost's injected identity headers via a templated respond — handle /whoami { templates; header Content-Type application/json; respond "{\"user\":\"{header.X-Authentik-Username}\",\"groups\":\"{header.X-Authentik-Groups}\"}" }. Spoof-proof because the proxy outpost strips client-supplied X-Authentik-* before injecting. app.js shows the admin button only when groups match homepage-admins; the /admin* 403 regex gate stays the real enforcement. Sequencing law learned the hard way: deploy caddy (whoami) BEFORE homepage — a homepage deploy whose e2e gate runs before /whoami exists fails the admin-visibility spec. Homepage card assets: every app card renders art (services.json field, hex-clipped .app-art); brand marks fetched from cdn.simpleicons.org/slug/ffb000 (dataset is lean — many slugs 404: openwebui, seerr, romm, seerr, loki, valkey, *arr beyond sonarr/radarr, tautulli, comfyui), improvised amber-hex monogram SVGs for the rest; official logos: openwebui.com/logo.png, seerr.dev/os_logo_filled.svg. Model cards render in registry order (zealot→dragoon→archon pinned in homepage_config.bats); e2e panel census counts #grid .panel .name, #grid .mi-name because model cards fuse into one panel. _(context: serverAppConfig homepage waves 2026-10-08 (commits through the icons/clickable/header wave); e2e gate 33/33 after caddy-before-homepage deploy ordering.)_
|
||||
- Dankclaw homepage admin identity-gate pattern (deployed 2026-10-08, serverAppConfigs): to gate a client-side control to authentik group members, add a Caddy handle /whoami { templates; header Content-Type application/json; respond "{...X-Authentik-Username/X-Authentik-Groups...}" } on the post-auth :8017 listener — the Authentik proxy outpost injects X-Authentik-* headers on its upstream request AND strips client-supplied copies, so echoing them back is spoof-proof. JS fetches /whoami and shows the control only when groups match (homepage-admins). The real gate stays server-side (the /admin* Caddy 403 on the same group header); the echo only controls what is shown. Sequencing law learned the hard way: deploy the caddy change (whoami endpoint) BEFORE the frontend that consumes it — the homepage deploy's e2e gate fired in the window where the new page expected /whoami and caddy didn't serve it yet, failing the admin-visibility spec. Related homepage conventions (all pinned in tests/homepage_config.bats): services.json registry order IS the card render order (model cards zealot→dragoon→archon); every app card carries hex-clipped art via an art field (brand marks from cdn.simpleicons.org/slug/ffb000, amber-tinted SVGs; improvised monogram SVGs = hex polygon + amber initials for apps with no mark — most *arr/wiki brands are absent from the CDN's dataset, check cdn 404s, don't assume slugs); card-level warn field renders as a red hover-label ("⚠ … — hover to read" with title tooltip); the common uncensored legal warning text (adult content, harmful/hateful/illegal, "sensitive offensive cybersecurity topics", "don't ruin this server for everybody") is shared byte-identical between dragoon and zealot. _(context: serverAppConfigs homepage wave 2026-10-08 (commit c1e192d): /whoami gate, app icons, warning unification; documented in serverAppConfigs-docs/Infrastructure/Homepage.md wave entries)_
|
||||
- Radiance (StillDeadcode) qwen4exp w4-GPTQ laws, measured 2026-10-07/08 on dankserver (Swift1.5 campaign): (1) ONE ENCODING PER EXPERT TABLE: rad_arch_declare rejects splitting ffn_gate_up_exps (w4) from ffn_down_exps (block-fp8) — "one table of experts takes one encoding, and only a whole expert kept plain bf16 is served beside it"; the PROTECTED section may hold only whole bf16 experts. (2) GPTQ w4 gate_up experts CANNOT be disk-tier streamed — stored form (u4[1x1] codes + fp8/f32 scale planes) is not the served form (dequantized tables); boot fails "DID NOT FIT ... store is not the container's"; they must fit the host pool (measured 11.72 GiB/rank for 48×512 at hidden 2560). (3) The activation arena scales HARD with --max-num-batched-tokens on w4 geometry: MNBT 32768 → 33.3 GiB arena on a 31.86 GiB card (262.5 MiB per 256 tokens); prefill-only jobs (KLD) run fine at MNBT 3072. (4) --kld-ref takes NO --kld-corpus (engine rejects the pair: tokens come from the reference); --kld-out FILE writes the JSON report + FILE.rows. (5) KLD scoring of a TP=2 container can place rank 1 on the host backend when the placement plan lets card 0 absorb the resident set (GPU[1] flat at 241 MiB while GPU[0] holds 25.88 GiB) — mathematically valid (reference-verified host kernels), and at pool 40960 it scored at 22.6 pos/s vs the disk-tier record's 3.3 pos/s. (6) rad-convert GPTQ on this box: ~6 min/layer × 49 layers (~4 h CPU, load 8), fully production-compatible (no GPU). _(context: serverAppConfigs swift15-radiance-w4dn8 campaign; candidate qwen4exp-swift15-w4dn8-i8-hc8m.rad 116,416 MiB)_
|
||||
- Dank-fleet family-facing model names (2026-10-08, Frank's StarCraft II Protoss theme): frank-large → frank-dragoon, frank-primary → frank-archon, frank-secondary → frank-zealot. These are DISPLAY names on the homepage and user docs only — endpoint services, OpenWebUI model ids, and API model ids remain frank-large/frank-primary/frank-secondary; agent configs (opencode/Pi) keep using frank-secondary. Supervision tiers (Frank's classification): frank-dragoon and frank-archon are safe for autonomous low-supervision use; frank-zealot (Qwen3.5-9B-Uncensored) requires tight human supervision. frank-zealot is the recommended compaction model for agent harnesses (compaction severely slows the bigger models far longer). Frank's rule: legal-consequence warnings for uncensored/abliterated models render as mouseover hover labels (title tooltips), NOT as always-visible wall text; other descriptive additions display inline. Homepage model cards fused into a single "Model Information" component (app.js renderModelInfo, .mi-* classes) — the e2e user-homepage census counts #grid .panel .name, #grid .mi-name because of this. A wip:"down" state renders the red "WIP: DOWN" badge (Seerr, Calibre-Web, RomM, Jellyfin are WIP: DOWN, Open links removed). _(context: serverAppConfig homepage Protoss wave 2026-10-08: /sfx/ ogg voices + unit art added; @data allowlist, deploy_homepage sfx block, caddy bats pin, e2e census fix (fused .mi-name cards) all required for a green deploy.)_
|
||||
- NEVER edit a bash script on disk while bash is executing it: bash reads scripts lazily by byte offset, so mid-run edits shift offsets and corrupt the remaining execution. On the 2026-10-07 Swift15 convert this made the plan "hot-patch the running convert script" impossible — the fix went to git (canonical copy) and the host file was left untouched until the run finished. Related corollary that bit the same night: a scp round-trip during an emergency edit silently dropped cleanup lines from scripts/convert-swift15-w4dn8.sh — after scp'ing a host file back as canonical, diff the whole file, not just the edited lines. _(context: serverAppConfigs swift15-radiance campaign: convert running from host copy; kld.jsonl cleanup lines vanished through the scp-back path; verified via grep on the running host copy (landmine disarmed by accident).)_
|
||||
- Radiance qwen4exp_fp8 plugin law (measured at rad-convert declare, 2026-10-07, Swift1.5 campaign): ONE ENCODING PER EXPERT TABLE per layer — "one table of experts takes one encoding, and only a whole expert kept plain bf16 is served beside it" (rad_arch_declare). A recipe may NOT split ffn_gate_up_exps (w4 GPTQ) from ffn_down_exps (block-fp8): the serve kernel is a single w4nl64a8h table for the whole expert set; the only legal beside-table content is whole experts cast dtype=bf16 (the PROTECTED section). Related: the ukisai/Swift1.5-Qwen3.8-Flash-Next "BF16" checkpoint is pure BF16 (1655 BF16 + 3 I64 PLE-index tensors, HF tensor names — no fp8 storage); rad-convert maps HF names to blk.N.ffn_*_exps.M itself. rad-convert's positional input must be the in-container path INCLUDING subdirs (basename-stripping the mount prefix breaks nested checkpoint dirs like /models/ukisai/name). GPTQ receipt line to watch in convert logs: "libquant: GPTQ quantises 'tensor' against N calibration file(s), K=cols". _(context: serverAppConfigs swift15-radiance-w4dn8 campaign, dankserver: the designed w4dn8 split tier was vetoed at rad-convert declare; recipe reduced to all-experts GPTQ w4 + 24 whole-bf16 protected experts (commit 73c37f4).)_
|
||||
- Dankclaw homepage family model names (2026-10-08, Frank's StarCraft Protoss theme): frank-large → frank-dragoon, frank-primary → frank-archon, frank-secondary → frank-zealot. Endpoint/API/OpenWebUI ids are UNCHANGED (frank-large/frank-primary/frank-secondary remain the ids agents and API configs must use). Frank's supervision classification: dragoon and archon are safe for autonomous low-supervision use; zealot requires tight human supervision. Compaction model for agent harnesses: frank-zealot (compaction jobs severely slow the bigger models). Frank's presentation preference: legal/uncensored warnings must be mouseover tooltip labels (⚠ hover-to-read), not visible wall text; other descriptive notes stay plainly visible. Media apps Seerr, Calibre-Web, RomM, Jellyfin are WIP: DOWN for family users (a third wip state "down" exists in services.json/badge CSS); Copyparty is the access point for all media apps' content including the WIP ones; Gitea and Plex use logins separate from the Authentik SSO; SearXNG is obfuscated-not-anonymous (upstream API keys tie to Frank's identity). Homepage assets: new top-level files (e.g. sfx/) must be added to the Caddy @data path allowlist AND the deploy_homepage copy list — same law as img/. _(context: serverAppConfig homepage Protoss rename wave, commit 203b0be; component = Model Information fused panel in shared/homepage/app.js with /sfx/ voices fetched from starcraft.fandom.com (MediaWiki API: action=parse wikitext, {{sm2|File.ogg}} quotes, imageinfo URLs for unit art).)_
|
||||
- Dankclaw homepage live verification (serverAppConfigs): (1) The post-auth Caddy listener (http://:8017, @home handler) serves files ONLY through the @data path allowlist (@data path /services.json /status.json /app.js /style.css /deployed.js /img/* in dankclaw/Caddyfile); anything unmatched falls to the catch-all and is answered with index.html as 200 text/html — a PNG arriving as text/html passes every status-code and file-listing check, so verify assets by content-type/bytes, never by HTTP status. New top-level homepage assets MUST be added to @data (pinned in tests/homepage_caddy.bats). (2) Live-check guarded pages with the eval browser runtime: browser.open({url, name}) (single options object; same for browser.close({name})); browser.tabs() lists {name,url,title}; tab.evaluate(string) runs JS in the page and returns a string — inside it, escaping: the eval script is built in a JS template literal, so use '\\n' (a raw \n interpolates and breaks page-code string literals); tab.url() returns an object, not a string. The shared headless browser keeps the qa-e2e SSO session, so https://www.franksdank.cloud/docs renders through Anubis+Authentik without manual login (the naked domain 308s to www and STRIPS the path — always probe www directly). Fetch-in-page probe: fetch('/img/x.png').then(r=r.status+' '+r.headers.get('content-type')+' '+blob.size) proved the @data bug; [...document.images].map(i=i.complete && i.naturalWidth0) proves rendering. _(context: 2026-10-08 homepage amber-theme session: @data allowlist bug (img/* served as index.html), eval-browser live verification of /docs and AI cards through the guard chain)_
|
||||
- Running-bash-script and scp round-trip discipline (serverAppConfigs campaign ops): (1) NEVER edit a bash script on the host while it is executing — bash reads script files lazily by byte offset, so edits shift offsets and corrupt the remaining execution. Fix committed copies, sync after the job ends. (2) After scp-based host edits, the HOST file is the ground truth for any running job (e.g. rad-convert loaded the recipe at launch); before git merges on the host worktree, diff each locally-modified/untracked file against the branch tip and commit the host copy as canonical when it is what the job loaded — a partial/stale copy can silently get committed (a committed recipe lost 10 of 24 protected experts this way; the running convert proved it). (3) The convert-script cleanup section (rm corpus files) was silently lost during a python line-rewrite round-trip: after scripted edits, grep for the intended deletions/markers to prove they survived. (4) A truncated ls | head -N listing caused a false "corpus deleted" alarm — when a critical file is absent from a truncated listing, stat the file directly before declaring loss. _(context: swift15-radiance-w4dn8 campaign 2026-10-07/08 on dankserver: KLD gate preparation between rad-convert stages)_
|
||||
- frank-large (Dankclaw halogen-flash-server) 400-zone law (2026-10-07/08): the harness clamps max_tokens = min(maxTokens, contextWindow − chars/4 est − 4096), and the est UNDER-counts the rendered prompt by ~3 % on code/agent traffic (measured twice: est 152,938 → server 157,388; est 158,906 → server 165,301). Compaction fires only at 85 % of the harness contextWindow, so window == engine ceiling made every request in the est ~146k–184k zone a guaranteed 400 ("max_tokens N does not fit: prompt is X tokens and the context is 200192"). Fix contract (deployed via scripts/frank-large.env): HALOGEN_CTX=262144 (native ceiling), HALOGEN_KV_POOL_POSITIONS=524288 (2×ctx, 14.0 GiB), models.yml frank-large contextWindow=217088 (= ceiling − 45k estimate headroom), maxTokens=45056; worst-case request ≈ 219k ≤ 262,144. Compaction input rides on frank-secondary: ≈ 170k ≤ its 200,704 slot. The general rule: a harness contextWindow must sit BELOW the engine ceiling by (maxTokens + reserve + est-error margin), never equal it. A "context is 200192" 400 after the env change means the OLD boot is still serving — deploy is pending, not the fix failing. _(context: serverAppConfigs frank-large token-limit fix session 2026-10-08: env+bats+AGENTS.md+runbook edited, frank_large_unit.bats 20/20 green, full suite + deploy dankclaw --service frank-large pending at hand-off. Supersedes the older "clamp max_tokens to 57344" note.)_
|
||||
- Dank fleet + omp harness token-fit law (diagnosed 2026-10-07, frank-large halogen 400s): the engine fit check is prompt_tokens + max_tokens ≤ engine window (halogen HALOGEN_CTX, llama.cpp slot ctx — a 400, never truncation), while the omp harness clamps max_tokens = min(maxTokens, contextWindow − chars/4_est − 4096) and only compacts at compaction.thresholdPercent (85 % in config.yml; reserveTokens-16384/70 % comments in models.yml/AGENTS.md were STALE). chars/4 under-counts the rendered prompt by ~3 % on code/agent traffic (measured: est 152,938 vs server 157,388 = 4,450 the 4,096 margin). Consequence: if harness contextWindow == engine ceiling, every request above ~4096/0.03 ≈ 137-146k est (65-85 % of a 200k window) is a guaranteed 400 — the zone between where the estimate error exceeds 4096 and the 85 % compaction threshold. FIX shape: engine ceiling − harness contextWindow must exceed maxTokens + ~10 % of contextWindow (safety for est error). Deployed fix for frank-large: HALOGEN_CTX 200192→262144 (native), HALOGEN_KV_POOL_POSITIONS=2×ctx=524288 (~28 KiB/pos = 14.0 GiB), models.yml contextWindow 217088 (212×1024) keeping maxTokens 45056 → compaction fires at 85 % = 184,524 est ≈ 190k measured (Frank's 180k target), worst-case request ≈ 219k ≤ 262,144 with 42k slack, and the compaction input (threshold − keepRecentTokens 20,000 ≈ 170k) still fits frank-secondary's 200,704 compaction slot. Frank explicitly demanded reaching 180k tokens on frank-large; he considers prior config "misunderstood" — check these three numbers (CTX, contextWindow, thresholdPercent) together, never one alone. _(context: serverAppConfig frank-large halogen-flash-server 400 "max_tokens does not fit" campaign, 2026-10-07; artifacts: scripts/frank-large.env, tests/frank_large_unit.bats fit-law test, models.yml frank-large block)_
|
||||
- OpenWebUI 0.6.x (open-webui on Dankclaw): external OpenAI-connection models are invisible to non-admin users unless a workspace row EXISTS in the model table (singular) with id == upstream model id, base_model_id NULL, is_active 1, PLUS an access_grant row (resource_type='model', resource_id=model id, principal_type='user', principal_id='*', permission='read'). The /api/models aggregation (utils/models.py get_filtered_models) DROPS models with no info for role=user; the workspace row supplies the info; the admin bypasses the filter, so admin UI looks fine while all family users see "No results found". An alias row (id != base_model_id) breaks chat completion ("Model not found") — id must equal the upstream id. API keys for regular users require user.permissions.features.api_keys=true in the config table (default false); the UI path is User menu → Settings → Account → API keys → Show → Create new secret key (0.6 layout, not the old SettingsAPI Keys tab). All three endpoints: DB at /home/dank/Containers/open-webui/data/webui.db (host), mount /app/backend/data; no sqlite3 CLI on host — use host python3; live process caches model lists ~1 s (MODELS_CACHE_TTL). The 2026-10-07 18:35 picker rebuild (frank-secondary swap) dropped the grants — fixed by inserting the row+grant pair.
|
||||
- Dankclaw Authentik (auth.franksdank.cloud) human roster as of 2026-10-08: akadmin (pk 4, superuser, frank8465@protonmail.com), admin (pk 5, admin@franksdank.cloud, homepage-admins only — legacy setup-era account, deactivation pending Frank's call), alex (pk 15), ridley (pk 13), kassie (pk 14). Non-human: AnonymousUser, qa-e2e, qa-e2e-admin, svc-jellyfin-ldap, ak-outpost-* (path goauthentik.io/outposts). The human-user template (documented in serverAppConfigs-docs/Infrastructure/Authentik.md §User provisioning playbook) = username + user@franksdank.cloud + display name + is_active + set_password + groups {Non-Admin Users, jellyfin_users, openwebui_users}; EXCLUDE authentik Admins / homepage-admins / *admins. Audit a user row BEFORE deleting: MFA device tables (authentik_stages_authenticator_{totp,webauthn,static,email,duo,sms}*device), authentik_core_token, app entitlements, authentik_providers_proxy_proxysession (stale claim cache). Event history survives a user delete (authentik_events_event "user" is denormalized JSONB, filter with "user"-'pk'='N'). Column names in the authentik-postgres DB differ from the ORM: certificatekeypair has no user_id, authentik_core_user has no source_id, events column is quoted "user" (jsonb) — select * / information_schema first, never guess. Passwords travel only on the ak shell stdin (heredoc over ssh), never argv/ps. _(context: 2026-10-08 session on Dankclaw: created kassie (pk 14), deleted pk 7 and recreated alex (pk 15) into the template, audited via ak shell + direct psql against authentik-postgres. Roster verified live with User.objects.exclude(path__startswith="goauthentik.io").)_
|
||||
- Dank fleet operational constraint (from Frank, 2026-10-07, verbatim intensity intact): "DO NOT FUCKING TOUCH ANYTHING BELONGING TO FRANK-PRIMARY" while dankserver's radiance box is running a multi-day quantization job (the Swift15-radiance KLD/calibration campaign). Practical meaning for an agent: do not restart, redeploy, benchmark, mask, flag-tune, or commit-and-push changes to scripts/frank-primary.env / frank-primary.service / serve-frank-primary.sh, and do not run the fleet-wide deploy.sh dankserver drain (it walks every service revision marker and would fire the frank-primary restart). Test failures that belong to frank-primary (e.g. tests/frank_primary_unit.bats asserting FRANK_PRIMARY_MAX_SEQS=2 while the env says 3) are NOT yours to fix during the window — commit only the files in your own task's blast radius; a green-suite gate yields to the campaign. Second signal from the same window: when Frank reports an endpoint healthy and says "no need to verify", stop benchmarking — record the evidence you already observed (boot ledger, e2e gate) and name the evidence gap explicitly instead of probing further. _(context: serverAppConfigs session 2026-10-07: frank-secondary (Dankclaw) swap work was blocked by a pre-existing frank-primary test failure; asking about resolving it drew the hands-off instruction. Deployed frank-secondary at ec6184d/5278f5b with frank-primary untouched.)_
|
||||
- Dank fleet (serverAppConfigs): every LLM serve script is a DIRECT systemd ExecStart (frank-primary/large/secondary), so rewriting one with a file-writing tool that drops #!/usr/bin/env bash makes the unit boot-loop with status=203/EXEC, Mem peak: ~1.9M, Failed at step EXEC spawning /home/dank/.../serve-svc.sh in the journal, and nvidia-smi flat at 1 MiB (the GPU is never touched). Diagnostic: 203/EXEC + zero GPU use = the script never ran; check head -1 and the exec bit (git ls-files -s shows the tracked mode; git preserves it, so a non-exec mode is committed drift). systemctl status's unit Description shows the LAST-deployed era's text, not the repo's — a stale Description is cosmetic until the next relink_unit (don't restart a serving endpoint for it). Related llama.cpp fact: the log line n_ctx_seq (N) n_ctx_train (M) is informational when N M (context below the trained window is safe); it is only alarming when a drafter runs ABOVE its own n_ctx_train. _(context: 2026-10-07 frank-secondary Qwen3.5-9B swap on Dankclaw: my rewrite of scripts/serve-frank-secondary.sh lost the shebang; the unit restart-looped 14 times (10 s apart) before the deploy's health gate; fix commit ec6184d. Boot ledger after the fix: 21,436 MiB used / 2,689 free, KV 12,544 MiB f16 K + f16 V.)_
|
||||
- llama.cpp master (build 5fc4f3c) attention_k_eq_v does NOT shrink the KV cache — the cache bills the K+V pair TWICE: 6.125 B/elem at f16 K + q8_0 V, 2.125 B/elem at q8_0/q8_0. Verified byte-exactly on Gemma4-26B-A4B (frank-secondary, dankclaw 3090, 2026-10-07) against the engine's own verbosity-4 llama_kv_cache lines: full cache = n_full_layers × n_head_kv × head_dim × rate × ctx = 5 × 2 × 256 × 6.125 × 200,704 = 3,001.25 MiB ✓; SWA cache = n_swa_layers × window × n_head_kv_swa × head_dim × rate = 25 × 1024 × 8 × 256 × 6.125 = 306.25 MiB ✓. Vcur = Kcur is a compute-graph mapping only. Also: read n_head_kv per layer (it is an ARRAY — gemma4 has 8 KV heads on sliding layers, 2 on full-attention layers) and sliding_window_pattern from the load log at verbosity ≥3, never from the HF config's single-value fields. Lesson: earlier fleet notes claimed the pool "bills K-only" (2.05 GiB at 200,704) — that figure came from arithmetic, not from the engine; a rate law must be fitted to the engine's own printed buffer sizes (-v/--log-verbosity 4 makes them appear in the journal). _(context: serverAppConfigs frank-secondary Gemma4 swap 2026-10-07: the f16-K raise (+2,161 MiB KV) booted at 21,868 used / 2,257 free because the ubatch 2048→1024 cut returned ~616 MiB of compute buffer; decode measured 139.9 t/s temp 0. Runbook: serverAppConfigs-docs/LLM/frank-secondary-llama-gemma4-q4km-200704.md)_
|
||||
- frank-secondary (Dankclaw RTX 3090, :7771) runs Gemma4-26B-A4B QAT Uncensored Q4_K_M (HauhauCS, 16,780,192,888 B) since 2026-10-07, ONE 200,704-token slot (--parallel 1), f16 K + q8_0 V, MTP OFF. Measured VRAM (nvidia-smi at boot, the referee, not arithmetic): MTP OFF = 21,890 MiB used / 2,235 free; MTP ON (separate gemma4-assistant sidecar mtp-gemma-4-26B-A4B-it.gguf, 0.252 GiB, loaded via --spec-draft-model + --spec-type draft-mtp) = 23,806 / 319 free. So the drafter costs ~1.9 GiB. Gemma4 Q8 with 2×200,704 is IMPOSSIBLE on the 24 GB card (weights alone ~26 GiB). Measured engine facts, not estimates: the assistant warns "n_ctx_seq (200704) n_ctx_train (131072)" (draft ctx is built at the target's full window, 53 % past the assistant's trained window) and declares shared-KV layers (main-model layers 0-3 sharing with 28/29); the main model's n_ctx_train is 262,144, so 200,704 is inside its trained window. With MTP off the main boot prints neither warning. Q4_K_M 1×200,704 with the MTP assistant is knife-edge; MTP off leaves ~2.2 GiB margin, which is what funds raising K to f16 (+2.05 GiB: 5 full-attention layers × 200,704 tok × 8 heads × 256 dim × 2 B). All three fleet endpoints (frank-primary, frank-large, frank-secondary) now use frank-secondary as compactionModel; compaction input fits its 200,704 window (~133k worst case) — a single --parallel 1 slot means compaction requests serialize behind live generation. _(context: serverAppConfig main branch, 2026-10-07 frank-secondary swap from Swift-1.5-Qwen3.8-27B RCO-IQ3_XXS (2×114,688) to Gemma4-26B-A4B Q4_K_M; VRAM numbers read from live nvidia-smi boots, warnings from journalctl -u frank-secondary.)_
|
||||
- Gemma4-26B-A4B Q4_K_M on the RTX 3090 (frank-secondary, dankclaw), MEASURED boot 2026-10-07 — supersedes the same-morning head-math estimate that said 1×200,704 does NOT fit: Q4_K_M 1×200,704 ctx with q8_0 K+V BOOTS with margin: 21,890 MiB used / 2,235 MiB free. With MTP assistant loaded it is 23,806 / 319 (drafter costs ~1.9 GiB — its KV is built against the target's full window). Key sizing facts: KV scales ONLY with the 5 full-attention layers of 30 (25 sliding-window layers at window 1024 cost a fixed 1024-token ring each), 8 KV heads × 256 dim; attention_k_eq_v=True makes Vcur=Kcur in the compute graph (gemma4.cpp: absent wv → Vcur = Kcur) and in llama.cpp build 5fc4f3c the KV cache honors k_eq_v (pool measured ~2 GiB at 200,704, matching K-only accounting), so bill K-only, not 2×. llama.cpp master build at commit 5fc4f3c accepts cache types: f32, f16, bf16, q8_0, q4_0, q4_1, iq4_nl, q5_0, q5_1 — NO fp8/f8 (user asked for "FP8 KV" → q8_0 is the FP8 class and is what ships). MTP for Gemma4 is a SEPARATE gguf (arch gemma4-assistant) loaded via --spec-draft-model + --spec-type draft-mtp; disabling = SPEC_TYPE=none which must strip BOTH flags (flags move as a set). The endpoint serves ONE 200,704 slot, MTP OFF, vision via publisher mmproj. _(context: serverAppConfig frank-secondary swap 2026-10-07: HauhauCS/Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-MTP Q4_K_M replaced Swift-1.5-27B at 2×114,688. Verified via llama-server argv, /v1/models (n_ctx 200704, ftype Q4_K_M), nvidia-smi. Corrects the earlier estimate memory entry that claimed 1×200,704 needed ~29 GiB.)_
|
||||
- llama.cpp build (commit 5fc4f3c, 2026-10-01) KV cache types: f32, f16, bf16, q8_0, q4_0, q4_1, iq4_nl, q5_0, q5_1. NO true fp8 KV dtype — q8_0 is the "FP8 class" (1.0625 B/elem, 8.5 effective bits). The grep -ac arch on the llama-server binary returns 0 for ALL arch names (including qwen3_5 which it's actively serving) — the arch strings are stored in a packed/constructed form that grep can't catch; use source registration (llama-arch.cpp) to verify arch support, not binary grep.
|
||||
- Gemma4-26B-A4B (google/gemma-4-26B-A4B-it) VRAM fit on the RTX 3090 (24,125 MiB), computed 2026-10-07 for the frank-secondary replacement question: Q4_K_M 2×200,704 does NOT fit — estimated ~33 GiB. Breakdown: weights 17,958 MiB (Q4_K_M 16,018 + mmproj-BF16 1,700 with compute buffers + MTP assistant 240) + KV pool 7,864 MiB (401,408 tok × 5 full-attn layers × 8 KV heads × 256 dim × 2 [K+V] × 1 B q8_0) + MTP draft KV ~1,600 MiB + compute/runtime overhead ~5,800 MiB. The swing factor: Gemma4 is MoE hybrid, 30 layers = 25 sliding-window (fixed 1024 ring) + 5 full-attention (layers [5,11,17,23,29]), so KV scales only with the 5 full layers, not 30. CRITICAL non-obvious build fact: in llama.cpp master build 5fc4f3c, attention_k_eq_v: True makes Vcur=Kcur in the compute graph (gemma4.cpp: Vcur = wv ? ... : Kcur), but the KV CACHE still allocates full 2× (K+V) — there is no k_only flag, so k_eq_v does NOT halve the pool. Feasible shapes: Q4_K_M 1×64k (~24.4 GiB, razor-thin) or Q3_K_M 1×128k (~24.1 GiB, needs a Q3_K_M download); dropping MTP saves ~1.6 GiB. Gemma4 MTP is a SEPARATE file (mtp-gemma-4-26B-A4B-it.gguf, arch "gemma4-assistant") loaded via --spec-type draft-mtp, unlike Swift's in-file MTP head. Diagnostic lesson: grep -ac arch llama-server returns 0 for EVERY arch name (including qwen3_5 which it's actively serving) because the packed binary doesn't expose arch names as plain strings — do NOT use grep to check arch support; check the source tree (src/models/gemma4.cpp exists at the build commit) instead. _(context: serverAppConfigs, frank-secondary (llama.cpp on RTX 3090, Dankclaw) — evaluating HauhauCS/Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-MTP as a replacement; build /home/dank/GitPackages/llama-master/build-cuda/bin/llama-server at commit 5fc4f3c.)_
|
||||
- Two durable lessons from the Swift15 campaign (2026-10-07, dankserver): (1) The script-swap law applies to plain scp too: overwriting a bash script a running process owns shifts its lazy-read byte offset — when the long child (podman run | tee) exits, bash resumes mid-token and dies rc=127 ("--security-opt: command not found") AFTER the real work succeeded. Recovery pattern: gate on the engine's own receipt (e.g. kld.json "rows": 50235 + file count + size) and resume the pipeline with a SKIP flag (kld-then-calib.sh SKIP_KLD=1); never restart the long job over a wrapper death. (2) serverAppConfigs docs seam "docs index links every docs page" (tests/docs_completeness.bats) requires every new page under serverAppConfigs-docs/ to be linked from the TOP-LEVEL serverAppConfigs-docs/index.md — a link in the sub-index (LLM/index.md) is not enough; add the top-level entry in the same commit as the page. _(context: serverAppConfig swift15-radiance-w4dn8 branch: KLD recording wrapper killed by mid-run scp of commit 45de2fd (rc=127, record intact); docs seam 135 failed until the new runbook was linked from the top-level index.)_
|
||||
- Overwriting a running bash script (scp/git checkout) shifts bash's lazy-read byte offset: the running script resumes mid-line at the next read boundary and executes garbage — symptom is a bogus "line N: token: command not found" and exit 127 AFTER the real work finished. Recovery pattern: judge the ARTIFACT, not the wrapper — e.g. radiance KLD record was complete (kld.json rows: 50235) despite the wrapper's rc=127; add a resume flag (SKIP_KLD=1) that gates on the artifact receipt (rows count, file count, size floor) and skips straight to the next phase. Prevention: campaign-named script files launched once per phase; never scp a script a running process owns; sync edits BEFORE launch. (serverAppConfigs 2026-10-07: mid-run scp of a perf edit killed the 4-hour KLD wrapper; KLD reference 24 GiB/50,235 rows survived.) _(context: Third fleet occurrence of the bash script-swap hazard (after the 2026-10-01 autoresearch collision); this one added the artifact-receipt resume pattern.)_
|
||||
- Radiance (StillDeadcode) VRAM ledger laws on dankserver R9700 (32 GiB card), measured 2026-10-06 bf16 disk-tier boots: the weights slab DERIVES as card − headroom − vram-kv-mib; the engine refuses (DID NOT FIT) and never shrinks for you. The disk tier has a hard weights floor of 9.57 GiB (4.21 GiB of two layer buffers + 5.36 GiB fixed weights) — so on a 32 GiB card with ~18 GiB fixed per-rank claims (QSA/DeltaNet/arena), --vram-kv-mib is capped near 3.5 GiB; 3072 MiB is the safe value, 4608 starves the disk tier. The ratio-derived pool (default EVC split) came up SHORT by 217 MiB at 32×16k fp8-KV slots — for offline (KLD/calib) shapes set --vram-kv-mib explicitly. Disk-tier prefill rate is STEP-SIZE-driven, not slot-count: MNBT 32768 warms to ~1,360 tok/s, MNBT 65536 ≈ 930 (vendor); slot count only queues pushes. Also: a foreground-podman systemd unit (frank-primary pattern) shows systemctl is-active=failed after every intentional stop (main process exits 143 on SIGTERM) — cosmetic, not a crash; check ExecStop status + journal, not is-active. _(context: serverAppConfig swift15-radiance-w4dn8 campaign 2026-10-06: three boot failures on the bf16 KLD/calib shape cost two relaunch cycles; the slab formula + disk-tier floor explain all three.)_
|
||||
- Radiance offline-run (KLD record / calibration serve) sizing law on dankserver, measured 2026-10-06 with the bf16 qwen4exp container: the VRAM weights slab DERIVES as card − headroom − KV, and --weights-disk-tier REFUSES under ~9.57 GiB (4.21 GiB layer buffers + 5.36 GiB fixed) — so on a 32 GiB card, explicit --vram-kv-mib must stay ≤ ~3.5 GiB for offline arms (3072 works with --max-num-seqs 4 at 16k fp8); the ratio-derived default came up SHORT by 217 MiB at 32×16k and the engine never shrinks for you. Recording is prefill/disk-bound, so fewer resident slots only queue pushes. Also: the image ENTRYPOINT is already radiance (passing "radiance" as argv → "E unknown option 'radiance'" rc=2), and the image has no ls — exec probes use sh. _(context: serverAppConfig swift15-radiance-w4dn8 campaign 2026-10-06: three failed KLD boots (entrypoint arg, KV SHORT, weights-slab starvation) before the offline shape booted; codified in scripts/run-swift15-bf16.sh comments.)_
|
||||
- Radiance (StillDeadcode) engine-feature probes MUST grep the architecture plugins, not just the binaries: model-specific features (calibration tap, KLD, QSA) live in /opt/radiance/share/radiance/architectures/arch.so (e.g. qwen4exp_fp8.so carries RADIANCE_CALIB_DIR/RADIANCE_CALIB_TOKENS on 1.2.0; the 1.0.13 plugin has RADIANCE_CALIB_DIR only) and quantizers in share/radiance/quantizers/libquant.so (gptq has calib=/damp/act_order/cd on BOTH 1.0.13 and 1.2.0). Grepping only bin/radiance produced a false "1.0.13 has no calibration tap" verdict. ALWAYS run a positive control (grep -c a string known present, e.g. kld-record) in the same probe; the image has NO strings(1) — use grep -ac on binaries. The radiance image ENTRYPOINT is already radiance: passing "radiance" as argv to podman run yields "E unknown option 'radiance'" rc=2. _(context: serverAppConfig swift15-radiance-w4dn8 campaign 2026-10-06: GPTQ Hessian pipeline version-blocker investigation on dankserver; two false negatives (missing strings tool; grepping the wrong file) cost a pin re-evaluation.)_
|
||||
- omp harness max_tokens clamp (pi-ai simple-options.js): max_tokens = min(models.yml maxTokens, contextWindow - estimateContextTokens - 4096). The 4096 CONTEXT_SAFETY_TOKENS is hardcoded in node_modules. estimateContextTokens = last assistant usage.totalTokens (actual from server) + Math.ceil(chars/4) for trailing messages after the last assistant turn. The chars/4 heuristic underestimates code-heavy content by ~2.8%, so in coding-agent sessions with tool results the estimate can be 4000+ tokens low, exceeding the 4096 safety buffer. Engines differ in how they handle over-cap max_tokens: Halogen (flash-server) returns 400 (max_tokens N does not fit: prompt is P and the context is C, leaving room for R), llama.cpp silently clamps. Fix: lower maxTokens in models.yml so the harness clamp stays under the engine's actual room even with estimation error. Halogen's max_tokens INCLUDES thinking tokens (no separate thinking budget field sent on the OpenAI path — thinkingTokenBudgetField is undefined for Halogen).
|
||||
- OMP harness max_tokens clamp (pi-ai/dist/api/simple-options.js): max_tokens = min(maxTokens, contextWindow − estimateContextTokens − 4096). estimateContextTokens = last assistant usage.totalTokens + Math.ceil(chars/4) for trailing messages (pi-ai/dist/utils/estimate.js). The chars/4 heuristic underestimates code-heavy content by ~2.8 %; CONTEXT_SAFETY_TOKENS = 4096 is hardcoded in node_modules (cannot be patched via config). Compaction threshold is contextWindow − reserveTokens (default 16,384 in config.yml), NOT the stale thresholdPercent: 70 comment in models.yml. Fix pattern for 400 "max_tokens N does not fit": lower maxTokens in models.yml to ≤ contextWindow − compactionThreshold − worstCaseEstimateError so max_tokens = maxTokens (fixed) rather than available (variable). Applied 2026-10-06: frank-large 57,344→45,056, frank-primary 36,864→16,384. _(context: serverAppConfig, 2026-10-06: diagnosed 400 "max_tokens 50013 does not fit" from frank-large; traced through pi-ai simple-options.js clamp + estimate.js chars/4 heuristic; fixed by lowering maxTokens in models.yml)_
|
||||
- Approved design (grilling session 2026-10-06, Frank's answers; execution pending — start with the purge + download + recipe file, service live). Goal: convert ukisai/Swift1.5-Qwen3.8-Flash-Next (335.3 GiB BF16) and serve it on frank-primary (radiance, 2×R9700 TP=2, dankserver:7770). Decisions: in-place swap of frank-primary (FP8-native .rad kept as rollback); engine pinned to the radiance 1.2.0 digest; keep the vendored Froggeric v22.5 template (verified: it handles vision + exposes tool_call_format) and sampler 1.0/0.95/20; base recipe = qwen4exp-w5dn64-i8-hc8m with the expert rules replaced by **down = rtn codes=fp8_e4m3 block=128x128 scale=bf16 and gate_up = gptq table=w4nl group=64 … cd=3 calib=$CALIB** (Frank's "down w8, rest w4"); everything else verbatim (i8 trunk, hc/mtp fp8 rows, ple_ngram fixed scale 1.99317932128906e-4, u2 draft head, QSA indexer + norms untouched). Expected container ≈130–140 GiB, experts ≈38 GiB/rank. Ten protected bf16 experts recomputed for Swift from per-expert down-Hessian energy + routing share from the engine's expert-selection metrics (even count per rank half). Quality gate: text+assistant mean KL ≤ 0.03 and top-1 ≥ 95 % (KLD corpus 1.8M tokens, source-labelled, rendered with Froggeric, disjoint from calibration); throughput gate decode ≥ 31.72 t/s × 0.97 and prefill ≥ −20 % vs the replicated baseline; on quality failure escalate gate_up w4→w5nl (pre-authorized, Grams reusable). Calibration corpus 10M agent-shaped tokens (45 % code / 30 % prose / 25 % tool JSON). Shape carried over: TP=2, wht6, seqs 3, maxlen 200000, MNBT exactly 2048 (MTP resonance law), KV fp8, spec 2, PLE in RAM, host pool 40960, with an in-window EVC bracket {0.82, 0.90, 0.95} because the smaller expert mass changes the residency optimum. Drain the whole campaign (multi-day, no clock). Disk: purge .AGENTTRASH glm-purge (172 GB) + v4flash-abandoned (42 GB) + models/stage-qwen4exp-fp8 (211 GB) + models/Qwen3.8-Flash-Next-Uncensored-FP8 (173 GB) + models/Qwen/Q _(context: serverAppConfig, frank-primary (radiance, dankserver). Grilling session produced a 25-decision design tree; Frank accepted every recommendation except Q4 (kept Froggeric over Swift's own template) and Q9/Q16 (chose the down=fp8/gate_up=w4 hybrid over the recommended all-fp8 experts). Confirmation to execute was pending when the session context ended; nothing on disk had been changed yet (working t)_
|
||||
- radiance (StillDeadcode) precision facts, verified in source at /tmp/rad-src (Codeberg main, 1.2.0): (1) There is NO servable 6-bit tier. codes= accepts fp6_e2m3/fp6_e3m2/u6/i6 and table= = {nf4, iq4nl, w4nl, w5nl, binary, ternary}, but the qwen4exp MoE block serves exactly kServed[8] = {fp8a8, w4a8, w4a8h, bf16, w4nla8h, w4nl64a8h, w4nl32a8h, w5nl64a8h} (arch/common/rad_block_moe_fp8.h:457); dense serves fp8a8/i8a8/fp8a16/mxfp4a8/bf16 + w2 draft head. "A container can hold only what some kernel reads" — rad-convert REFUSES an encoding no kernel reads (docs/TOOLS.md:184), so a q6 recipe fails at convert, not at boot. (2) Published Flash-Next quality ladder (docs/MOE-W4.md:24-32, mean KL / top-1, all tokens): w4nl64a8h served 0.0789/92.85%, w5dn64 0.0717/93.34%, bf16 floor 0.0399/95.13%; text+assistant scope w5dn64 = 0.0204/95.10%. Any KL gate ≤0.05 on the all-token scope is unreachable by every quantized tier. Author's lever (MOE-W4:68): 5 bits on layers 16-31 both projections, ~0.066 expected. Heavier expert tiers cost PREFILL (w5dn64: decode +4%, prefill −19%). (3) Calibration: the ENGINE records — serve with RADIANCE_CALIB_DIR [+ RADIANCE_CALIB_TOKENS, default 65536], drive prefill, it writes gram.rrank.weight stem.bin once. A BF16 Flash-Next (335 GiB) serves on two cards via --weights-disk-tier and records ~930 tok/s (10M in 3 h) — documented verbatim (MOE-W4:244-253, GUIDE:529). down gets a per-expert Hessian; gate_up is pooled per layer. GPTQ adds "tens of minutes" per 48-layer container + 1.3 GB factor cache. (4) Scoring: radiance --model BF16.rad --kld-record DIR --kld-corpus C.jsonl once (exact wire + bf16 KV), then --kld-ref DIR --kld-out REPORT.json per candidate; report is per source, so turn-type scopes come from labelling the corpus's source field; corpus must be disjoint from the calibration corpus; speculation/prefix-cache off in that mode, MNBT ≤ ~8192 (logits buffer 2 GiB/rank per 4096 tokens). (5) ukisai/Swift1.5-Qwen3.8-Flash-Next: 335.3 GiB BF16 _(context: Grilling session 2026-10-06 on serverAppConfig: designing a rad-convert recipe for ukisai/Swift1.5-Qwen3.8-Flash-Next on frank-primary (radiance, 2×R9700 TP=2). Facts gathered from a shallow clone of Codeberg StillDeadcode/radiance + HF API/safetensors-header probes; a scout subagent died on frank-secondary's 131k slot overflow (exceed_context_size_error) so the facts were collected inline.)_
|
||||
- radiance (StillDeadcode) serve-side precision surface, verified from source (clone of main, 2026-10-06) — the quantizer can WRITE more widths than any kernel can RUN, and rad-convert enforces it: "A container can hold only what some kernel reads. The converter checks every planned encoding against the layout hooks of the kernels declare resolved and refuses one nothing accepts" (docs/TOOLS.md:184-186). Consequences: (1) NO 6-bit tier exists — codes= accepts fp6_e2m3/fp6_e3m2/u6/i6 (libquant/lq_quant.h:15-17) and RAD_FP6E2M3 is a real ABI dtype (abi/rad_types.h:186), but libr4d has zero fp6 GEMM rows (lq_common.h:105 calls the fp6 grids conversion-only) and table= codebooks stop at five bits: nf4, iq4nl, w4nl, w5nl, binary, ternary (lq_quant.h:19) → a q6/fp6 recipe FAILS at rad-convert, not at boot. (2) The qwen4exp MoE block serves exactly 8 formats, arch/common/rad_block_moe_fp8.h:457: kServed = fp8a8, w4a8, w4a8h, bf16, w4nla8h, w4nl64a8h, w4nl32a8h, w5nl64a8h — anything else is RAD_E_DTYPE (error names the list, :671); dtype is chosen PER PROJECTION PER LAYER (:396), so fp8 down + w5nl gate_up in one container is legal. (3) Dense/trunk GEMMs serve bf16, fp8a8, fp8a16 (m1/mt), i8a8, w4a16/w4a8(+tiled/prefill/asym), w8a16/w8a8, mxfp4a8, and w2a8 for the draft head (libr4d/r4d.h:430-670, r4d_moe.hip:2393-2527). (4) Calibration has no recorder tool: the ENGINE records — serve with RADIANCE_CALIB_DIR (+ RADIANCE_CALIB_TOKENS, default 65536), drive prefill traffic, it writes per-layer Grams and prints "radiance: calibration wrote layer Hessians after N tokens to dir"; calib= is accepted by gptq/awq/autoround/paroquant, unset $CALIB is refused, a weight with no Gram is quantised uncalibrated and says so on stderr (docs/MOE-W4.md:229-271, TOOLS.md:158). (5) A BF16 Qwen3.8-Flash-Next checkpoint (335 GiB) can be SERVED on two gfx1201 cards via --weights-disk-tier (~28 s a step, ~930 tok/s recorded, 10M tokens ≈ 3 h) precisely to record bf16-domain Hessians for a GPTQ recip _(context: serverAppConfig, 2026-10-06: grilling session for converting ukisai/Swift1.5-Qwen3.8-Flash-Next to radiance with a qwen4exp recipe; asked to "bump q4/q5 to q6/fp6" and the kernel library proved that impossible. Facts from /tmp/rad-src (clone of Codeberg StillDeadcode/radiance main) corroborated by a scout run.)_
|
||||
- Radiance (StillDeadcode) precision facts — primary-source research 2026-10-06, not yet in local docs. PROJECT MAP: codeberg.org/StillDeadcode/radiance = the standalone C++/HIP engine (what frank-primary runs; latest 1.2.0, 2026-10-06); codeberg.org/StillDeadcode/vllm-radiance = DEPRECATED vLLM wrapper ("I ditched vLLM… it is faster, easier to work on, starts faster and has less cpu overhead"); codeberg.org/StillDeadcode/libr4d = shared kernel library, entry points named for exact geometry and rejecting mismatch (attn_decode_h256_gqa6_fp8kv, gdn_chunk_scan_k128_v128_c64_bf16, ar_oneshot_2rank_exact). PRECISION LADDER: block-FP8 (E4M3 codes + bf16 block scales) is the native tuned path — the "preshuffled block-FP8 dispatcher"; replacing it with generic upstream AITER linear is 9–11 % slower in matched controls, and every upstream-qualified model is FP8 (Qwen3.8-27B-FP8, Qwen3.6-27B-FP8, Qwen3.6-35B-A3B-FP8, Gemma-4-31B-it-FP8). MXFP4/W4A8 (AMD Quark weight path) stores packed OCP group-32 MXFP4 weights "and dynamically quantizes activations to FP8 E4M3 so the kernels use RDNA4's native FP8 WMMA path" — compute is FP8 regardless of weight packing. bf16 = KV (auto), skinny GEMMs, GDN (gdn_chunk_scan_…_bf16). INT4/INT8 = draft head (decoder projections "packed to signed symmetric int4… 4.25 bits per weight", int8 kernel above 16 rows). HARDWARE "WHY": RDNA4 WMMA is integer-biased — "gfx1201's f16 matrix instruction is half the rate of its int8 one and a quarter of its int4 one" ⇒ int4 int8 f16; FP16 saturates the 128-bit memory interface natively while FP8/INT8 need a fused double-K WMMA to reach 128-bit loads (AMD GPUOpen WMMA guide part 2). Sources: codeberg StillDeadcode/radiance + /libr4d + /vllm-radiance READMEs, github.com/magiccodingman/vllm-radiance fork README, gpuopen.com/learn/wmma-guide-amd-rdna-4-gpus-part-2. _(context: Online research for the radiance precision question (serverAppConfig, 2026-10-06); findings are pending a docs fold-in and feed the Swift1.5-Qwen3.8-Flash-Next rad-convert recipe design.)_
|
||||
- Radiance (StillDeadcode) tensor-precision picture — verified 2026-10-06 from primary sources. Project map: codeberg.org/StillDeadcode/radiance = the STANDALONE engine (C++/HIP, no Python) that frank-primary runs; codeberg.org/StillDeadcode/vllm-radiance = DEPRECATED vLLM wrapper ("I ditched vLLM… standalone is faster, starts faster, less CPU overhead"); codeberg.org/StillDeadcode/libr4d = shared kernel library, entry points named for exact geometry (e.g. attn_decode_h256_gqa6_fp8kv, gdn_chunk_scan_k128_v128_c64_bf16, ar_oneshot_2rank_exact) and it rejects a shape mismatch rather than running. Precision ladder: (1) block-FP8 (E4M3 codes + bf16 per-block scales) is the NATIVE tuned GEMM path = the "preshuffled block-FP8 dispatcher"; replacing it with generic upstream AITER linear = 9-11% slower in matched controls; all upstream-qualified models are FP8 (Qwen3.8-27B-FP8, Qwen3.6-27B-FP8, Qwen3.6-35B-A3B-FP8, Gemma-4-31B-it-FP8). (2) Compute is FP8 WMMA on RDNA4's native matrix core — even the MXFP4/W4A8 (AMD Quark) weight path DYNAMICALLY quantizes activations to FP8 E4M3 to hit that native FP8 WMMA path (so compute is FP8 regardless of weight packing; MXFP4 = packed OCP group-32 weights). (3) bf16 = KV cache (the auto/bf16 option; fp8 is the qualified mandatory default), skinny GEMMs (split-K for small projections), and GDN (gdn_chunk_scan…_bf16, an fp16 matrix-core WMMA gated-delta-net path). (4) INT4/INT8 = draft head (drafter decoder projections "packed to signed symmetric int4… 4.25 bits per weight", int8 kernel above 16 rows). HARDWARE "why" (RDNA4 WMMA, gpuopen.com/learn/wmma-guide-amd-rdna-4-gpus-part-2): the matrix core is integer-biased — "f16 matrix instruction is HALF the rate of int8 and a QUARTER of int4", so hardware ranking int4 int8 f16; FP16 saturates the 128-bit memory interface natively but FP8/INT8 only do 64-bit loads, so the tuned GEMM kernels FUSE two WMMA instructions (double-K / K-dimension) to enable 128-bit vector loads and hit peak bandwidt _(context: serverAppConfigs LLM fleet — answering "what tensor precisions does the radiance runtime like"; online research against StillDeadcode Codeberg repos + AMD GPUOpen WMMA guide.)_
|
||||
- radiance runtime (StillDeadcode C++/HIP engine, libr4d kernels for gfx1201) tensor-precision support: FIRST-CLASS/native = grouped/block-FP8 (E4M3 codes + bf16 per-block scale planes) — the ACTIVE frank-primary serves the official Qwen3.8-Flash-Next-FP8 natively with no requantisation (the qwen4exp_fp8 plugin is built around it). The vllm-radiance wrapper line (frank-medium, frank-primary R9V) extends the same libr4d kernels to a quant ladder: MXFP4/MXFP6 (ParoQuant W4A8, e4m3fn scale, group_size=128, krot=8), Q8_0, IQ4_XS. KV cache: fp8 native (7.32 KiB/token on frank-primary) or bf16 unquantized (--kv-cache-dtype auto). Scales: bf16 (block-FP8 planes, PLE one-element scale). Speculation draft heads: blockfp8 (MTP/DFlash2). TP wire: wht6 = lossy FP8-class fused one-shot all-reduce (vs --tp-wire exact bit-exact). The distinction that matters: "radiance" (pure engine, no Python, native block-FP8) vs "vllm-radiance" (vLLM front end over the same kernels, the quant ladder). Sources: serverAppConfigs-docs/LLM/frank-primary-radiance-qwen38-native-fp8.md, frank-medium-radiance-mxfp4.md, frank-primary-r9v-radiance-qwen38-q8.md, RDNA4 Container Serving.md §1. _(context: Answered "what tensor precisions does radiance runtime like?" from the serverAppConfig LLM runbooks.)_
|
||||
- serverAppConfigs shared-checkout diagnostic (proven 2026-10-06): to determine the TRUE position of a dirty shared checkout, do NOT trust the abbreviated git reflog tail — it can show commits (e.g. c80c250) as recent HEAD@{N} that are actually ORPHANED (the branch was reset away from them; they're reachable only via reflog). Verify with three agreeing sources: git rev-parse HEAD, git for-each-ref refs/heads/branch, and git log --oneline (tip). In the 2026-10-06 case, the reflog tail showed ...81e6122 - 95923d3 - df6b5bb - c80c250 commit as the last entry, but HEAD+branch+log all sat at 81e6122 — the three bench commits were orphaned after an autoresearch sweep (final commit c80c250 concluded "KEEPING Frank's config", so the branch was reset back to the harness commit 81e6122). Also: an UNSTAGED, uncommitted deletion of a tracked file that is present on the target branch is LOSSLESS to git restore -- file (it's pure working-tree state, in no commit); confirm with git ls-files --stage file (tracked?) + git ls-tree branch -- file (present on target?). This is the safe "clean up the working tree" move before git checkout branch. _(context: Cleaning up the serverAppConfigs shared checkout and returning to main (2026-10-06): resolved an apparent reflog/log discrepancy on the autoresearch/...frank-pri-20261006 branch, restored an unstaged deletion of autoresearch.sh, and switched to main.)_
|
||||
- Ops laws confirmed live 2026-10-06 on dankserver: (1) Fleet containers are ROOTLESS under user dank — sudo podman ps/logs queries the empty ROOT store and shows NOTHING; production appeared "gone" while radiance was serving fine. Always query podman ps / podman logs AS dank without sudo; verify liveness via ss -ltnp (radiance pid on :7770) or /health. (2) frank-primary logging trap CONFIRMED: scripts/serve-frank-primary.sh runs podman run -d ... "$LOG", so /home/dank/logs/frank_primary.log holds 65 bytes (one container-ID hex line) and the boot ledger lives ONLY in podman logs frank-primary (rootless, as dank). Fix = foreground podman run under Type=simple (pattern proven on frank-large, commit ba67ca5); queued as post-segment maintenance commit (serve script is off-limits inside the autoresearch segment). (3) The radiance engine logs NO per-request draft acceptance server-side; timings.draft_n/draft_n_accepted are returned to the CLIENT only — any real-traffic spec-depth A/B needs a client-side collector. _(context: 2026-10-06 dankserver: production health checks during the autoresearch segment; root-store vs rootless-store confusion cost several turns; frank_primary.log inspected directly)_
|
||||
- frank-primary (radiance, 2xR9700 TP=2) decode-sweep results 2026-10-06 (21 runs): final config = Frank's original spec=3/seqs=2/maxlen=200k (commit c80c250); baseline 31.72 t/s at 43.6k depth replicated 31.72/31.74. Noise floor ±3% decode (identical config drew 33.06/32.03/30.90 — single-run deltas 3% are noise; NEVER keep on 3%); prefill noise ±2%. MNBT RESONANCE LAW: with MTP on, MNBT=2048 is a sharp resonance point, NOT an interior optimum — 2560 collapses decode to 2.6–7.7 t/s (acceptance stays NORMAL 0.82, verification stalls ~130ms/step, output truncates 213/256; reproduced on clean boot), 1536 truncates 57/256 + prefill −22%, 4096 won't boot at seqs=3. MNBT must stay exactly 2048 at this shape; mechanism unresolved (kernel boundary × 3 slots × MTP verify). HOST_POOL FLOOR LAW: 36864 MiB crashes boot; 40960 is the load-bearing floor of expert page-staging (49152 boots flat — not a slack knob). Real regressions only: spec=4 (−23%, acceptance 0.50), spec=0 no-MTP (−7%, MTP genuinely pays), EVC 0.84 (−8.9%). Flag surface 100% bracket-mapped; spec 1–3 × seqs 2–3 statistically tied; depth 3 kept on acceptance tiebreak (0.831 vs 0.755) — the 256-token greedy probe cannot express it; real-traffic A/B is the only separating instrument. _(context: frank-primary radiance autoresearch segment 2026-10-06 (21 bench runs, branch autoresearch/...frank-pri-20261006, final commit c80c250); harness: autoresearch.sh + scripts/autoresearch-frank-primary.sh)_
|
||||
- frank-primary (radiance, dankserver) tuning laws from the 2026-10-06 autoresearch window (branch autoresearch/iteratively-improve-the-performance-of-frank-pri-20261006, final config commit c80c250): (1) MNBT RESONANCE LAW — at MTP-on shapes, --max-num-batched-tokens=2048 is a sharp resonance point, NOT an interior optimum: 2560 collapses decode to 2.6-7.7 t/s (reproduced on a clean boot; draft acceptance stays normal at 0.82 while verify stalls ~130ms/step; generation truncated 213/256), 1536 truncates at 57/256 with prefill -22%, 4096 won't boot at seqs=3. MNBT must stay exactly 2048; re-testing is forbidden, mechanism unresolved (kernel boundary interaction between chunk size, slots, MTP verify budget). (2) HOST_POOL_MIB 40960 is the LOAD-BEARING FLOOR of the expert page-staging reservation — 36864 crashes boot, 49152 boots flat; it is not a slack knob. (3) MEASUREMENT LAW — decode noise floor is ±3% on identical config (33.06/32.03/30.90 on one config), prefill ±2%; single-run deltas 3% are noise, replicate before keeping any change. (4) Decode surface is bracket-complete: spec depth 1-3 statistically tied (0 and 4 are real regressions: 29.57 and 24.57), seqs 2-3 tied, expert-vs-cache 0.82 bracket-confirmed (0.78 crash / 0.84 -8.9%), maxlen 262144 steals resident-weights VRAM (-2.5%), GPU-headroom 48 and host-pool 49152 flat. Tie-break that kept Frank's original spec=3/seqs=2: draft acceptance 0.831 vs 0.755 favors real multi-turn traffic; the 256-token greedy probe cannot express it. The separating instrument for depth 2 vs 3 would be real-traffic per-request draft acceptance from the engine ledger (needs production window at depth 2, Frank's sign-off). _(context: serverAppConfigs autoresearch segment 2026-10-06 (frank-primary decode tuning, 21 runs): sweep closed at a tie with Frank's original config kept; MNBT resonance and host-pool floor laws measured; noise-floor law established by identical-config replication.)_
|
||||
- frank-primary radiance tuning window (2026-10-06, 20 autoresearch runs): (1) MNBT RESONANCE LAW — at MTP-on/seqs=3, --max-num-batched-tokens is a sharp resonance point at 2048, NOT a smooth curve: 2560 collapses decode to 2.6-7.7 t/s with normal draft acceptance (0.82) and verify stalling ~130ms/step, generation truncating at 213/256; 1536 truncates at 57/256 and loses 22% prefill; 4096 fails boot at seqs=3 (booted at seqs=2, only -2.6% there). MNBT must stay =2048 at this shape; mechanism unresolved. (2) HOST_POOL_MIB 40960 is the load-bearing FLOOR, not slack: 36864 fails boot, 49152 boots but is decode-flat. (3) Noise floor law: decode is ~±3% run-to-run on IDENTICAL config (33.06/32.03/30.90 measured), prefill ~±2% — single-run deltas under those bounds are not evidence. (4) Viable envelope at seqs=3: EXPERT_VS_CACHE 0.82 (0.78 crash / 0.84 -9%), MNBT 2048, TP_WIRE wht6 (bf16 crashes), HOST_POOL 40960. Spec depth 1-3 statistically tied (0 and 4 real regressions: 29.57 / 24.57). Endpoint was already well-tuned; no flag clears noise. Scratch-bench hazard: if a prior bench is abandoned mid-run, the next scratch boot may race the leftover engine — verify the previous arbench container is gone and prod restored before trusting a collapse measurement (one 92% collapse was initially suspect for this; MNBT=2560 collapse then reproduced clean-boot, confirming it real). _(context: Autoresearch segment on serverAppConfig, frank-primary (radiance, Qwen3.8-Flash-Next-FP8, 2xR9700 TP=2, seqs=3, maxlen=200k, spec=2): harness scripts/autoresearch-frank-primary.sh boots a scratch radiance container per run; 20/30 runs consumed, flag surface 100% bracket-mapped, no keep beyond noise. Frank's pending decision: keep spec=2/seqs=3 vs revert to his spec=3/seqs=2.)_
|
||||
- Single-run A/B on the radiance /v1/completions timings (256-token greedy decode, 43.6k-token prompt, fresh scratch container per boot on dankserver) has a true noise floor of ~±3% decode t/s, NOT the ±0.3% one might assume from short-interval probes. Identical config measured 33.06 / 32.03 / 30.90 t/s. Consequence: never keep a tuning change on a 3% single-run delta; re-run promising keeps once before believing them (and flag the first lucky draw as suspect), and bracket both sides of a tuned value before trusting an optimum. Boot-failure arms (health timeout ~720s, ~12 min) are informative walls, not wasted runs. Real, outside-noise effects at frank-primary: SPEC_TOKENS=4 (-23%, acceptance 0.83-0.50), SPEC_TOKENS=0 (-7%, MTP always pays), EXPERT_VS_CACHE=0.84 (-9%), KV shape seqs=4/maxlen=100k (-11%), and three boot crashes that define the seqs=3-viable envelope (EXPERT_VS_CACHE 0.78, MNBT 4096, TP_WIRE bf16 all crash at seqs=3; the wht6/2048/0.82 trio is what makes it fit). _(context: Autoresearch session on serverAppConfig (frank-primary radiance decode tuning, 2026-10-06): 16-run ladder over one flag at a time; runs 39/40/49 measured the SAME committed config (spec=2, seqs=3, maxlen=200k) at 33.06 / 32.03 / 30.90 t/s, exposing the true run-to-run spread; run #39 had been kept as "+4.2%" and had to be flagged suspect after the replication run exposed the draw.)_
|
||||
- frank-primary (radiance, TP=2 Qwen3.8-Flash-Next-FP8) decode tuning laws, measured 2026-10-06 on a deterministic 43.6k-token greedy probe (scripts/autoresearch-frank-primary.sh, decode_tps = timings.predicted_per_second, noise ~0.3%): (1) MTP depth is UNIMODAL at 2: depth1 31.89, depth2 32.40, depth3 31.72, depth4 24.57 t/s (acceptance collapses 0.83-0.50 at depth 4). SPEC_TOKENS=2 is the production value. (2) MAX_SEQS 2-3 (+2.0%, co-scheduling headroom; single-stream probe sees it via scheduler shape) wins; 4 regresses (-3.1%). (3) The KV-shape surface is peaked at seqs=3/maxlen=200000: every window mutation regresses depth decode (native 262144: -2.5%; 100000x4: -10.6%) because the KV pool trades against resident-weights VRAM at expert_tiered. (4) EXPERT_VS_CACHE=0.82 is bracket-confirmed optimal: 0.78 fails boot at seqs=3 (KV unallocatable), 0.84 costs -8.9%, 0.87 known collapse. (5) At seqs=3 the VRAM envelope is tight: MNBT 4096 and TP_WIRE bf16 both FAIL BOOT (720s health timeout) — wht6 + MNBT 2048 are what make the seqs=3 shape fit. (6) GPU_HEADROOM 48 and HOST_POOL 49152 are decode-neutral (the pinned-pool streaming is not the single-stream bottleneck). Net kept gain: 31.72 - 33.06 t/s (+4.2%), committed as spec=2 + seqs=3. _(context: Autoresearch session frank-primary-decode-perf, runs #35-#48, autoresearch branch, harness stop-prod/boot-scratch-arbench/bench/restore on dankserver)_
|
||||
- serverAppConfigs deployment-completeness policy: a behavior-affecting deploy needs the bats suite (./tests/run.sh) AND the Playwright e2e gate (tests/e2e/run.sh). deploy.sh normally runs the e2e gate as its [3/3] step, but its whole-tree clean-tree gate (scripts/deploy.sh lines ~116-120: git status --porcelain non-empty → FAIL) is blocked by ANY dirty file, including a parallel writer's unrelated WIP. When that happens, run the two gates standalone — ./tests/run.sh for the seams and ./tests/e2e/run.sh for the live-fleet e2e (it needs tests/e2e/.env.local credentials, drives the live fleet, and does NOT require a clean tree). This satisfies the deployment-completeness requirement without deploying; the actual service dispatch can be done manually (git push + remote ff-pull + systemctl start) when it only touches files the writer isn't touching. Proven 2026-10-06 on the frank-primary rad-tune/MNBT window: writer's dirty frank-large WIP blocked deploy.sh, so I ran tests/run.sh (all green) + tests/e2e/run.sh (33 passed) directly and restored frank-primary by manual push+pull+start.
|
||||
- radiance (StillDeadcode) rad-tune on this fleet: the kernel-selection cache is written to $RADIANCE_HOME/tune/machine.tune, and $RADIANCE_HOME defaults to /opt/radiance — which lives INSIDE the image, so a --rm container's cache dies with the container. To persist it you must mount a host dir over it (-v hostdir:/opt/radiance/share/radiance/tune) in BOTH the rad-tune run AND the serve run, and pre-create the host dir (the engine's create is NOT recursive). Machine key on this box: gfx1201-x3-AMD_Radeon_Graphics (3 devices: 2x gfx1201 R9700 + 1x gfx1036 iGPU). Declare at the shapes you serve: --tp 2 (a world-size-1 declare measures a program a TP=2 serve never runs) and --num-speculative-tokens 3 (puts the drafter's kernels in the measured graph; default is the container's own depth). On this build the two attention kernels (attn_paged, attn_paged_gate_quant) are SKIPPED by rad-tune ("no shape recipe; see tools/opshapes.h") — only the 36 gemm_nt (kernel,shape) pairs get tuned, so the cache covers GEMM tile selection, not attention. rad-tune is ~110-150s for this 173 GiB model at TP=2. The radiance image has NO bash and NO env/ls — only sh; invoke it as podman run --entrypoint sh image -c "cmd" (the ---before-args form breaks). _(context: frank-primary radiance cutover bench window 2026-10-06: rad-tune --tp 2 --num-speculative-tokens 3 on qwen4exp-fp8-native.rad; cache path + mount law + attention-skip observed live)_
|
||||
- Bonsai-2-27B swap on frank-small: scouted + smoke-booted 2026-10-06, then rolled back to Swift RCO-IQ3_XXS by Frank's order BEFORE any repo edit (tree stayed at bb3532d). Payload parked on disk for a retry. (1) RUNTIME LOCK: stock llama.cpp CANNOT load PQ2_0/PTQ1_0 ternary files (unknown types; Q2_0 loads silently-garbage — no Hadamard activation runtime). Required runtime = PrismML-Eng/llama.cpp fork (prism branch). Deployed prebuilt: release prism-b10754-2459f68 asset bin-linux-cuda-12.4-x64 (matches Dankclaw nvcc 12.4; CUDA 13.3 builds documented crash-prone; tarball sha256 80185ce29bed2456f099903a41fe20452121af0c2da6be84315c9145e46e3ad2), extracted FLAT — binary at /home/dank/GitPackages/llama-prism/llama-server, NO bin/ sublevel (a bin/llama-server path fails). Post-2026-09-23 releases carry the AVX-512 load-segfault fix #245 (Dankclaw Strix Halo = Zen 5 is the affected class; a pre-b10735 release would segfault). (2) PAYLOAD: dealignai/Bonsai-2-27B-Ternary-CRACK-GGUF Bonsai-2-27B-PQ2_0-CRACK.gguf = 7,206,168,928 B, sha256 5b24ea3eebc3e0bccd05fb474eb88b10c57699d71a5db2f29485e3789a70d55d; PrismML ternary {−1,0,+1} g128 in a Hadamard-rotated basis at 2.13 bpw of Qwen3.8-27B, abliterated at weight level (0/320 HarmBench-320 refusals vs base 299/320; MMLU −0.62 pp); 64-block qwen3_5 hybrid — NO MTP head in the file (draft-mtp refused) and ngram-* spec types have NO effect on this model (fork issue #203) ⇒ spec=none is the only contract; mmproj ships in the BASE repo prism-ml/Ternary-Bonsai-2-27B-gguf (mmproj-Q8_0 629,246,976 B, sha256 6807ede61d570bb86ba34b756a0fa109edc33668604de867c6ea6d8f1d631903; BF16 reference pack 931,145,856 B) at /home/dank/models/Bonsai2-CRACK/. (3) SMOKE MEASUREMENTS (2×131072 q8_0 KV, ubatch 512, SPEC none): boot ~2.5 s, VRAM 17,006 MiB used / 7,570 free (widest fit the endpoint has ever had — a third slot plausibly fits), n_slots=2 confirmed, vision projector loads. (4) FORK QUIRKS for the eventual swap: reasoning_effort "high" → HTTP 50 _(context: 2026-10-06 frank-small Bonsai-2-27B session: Hub scouted (both repos), all three artifacts fetched+sha-verified on Dankclaw, smoke boot on :7771 verified 17.0/7.6 GiB, then Frank ordered rollback to the prior Swift endpoint; payloads and fork binary remain on disk with no repo changes.)_
|
||||
- frank-small RCO-IQ3_XXS two-slot MTP-on pricing (2026-10-06): the MTP draft ctx measured ~2.4 GiB, not the ~0.4 GiB estimated from proposal depth — the draft context is built against the target's FULL n_ctx (262,144), per the existing "drafter's KV scales with the target's context" law, so price MTP draft ctx from a boot pair, never from the draft depth. Fit: 23,636 used / 489 MiB free (tightest frank-small fit in history; the 2–3 GiB floor knowingly violated by Frank's explicit order). Two companion facts for this payload/build: (1) the UNCENSORED model thinks more than censored Swift 1.5 — a 64-token vision probe came back all reasoning_content with empty content (finish_reason length); the probe needs a 128-token budget. (2) The current llama.cpp master /completion timings object exposes draft_n, draft_n_accepted, predicted_per_second — NOT n_prompt/n_text/t_prompt/t_text — so decode rates come from predicted_per_second and acceptance from draft_n_accepted/draft_n (measured 49.8–62.4% on this quant). _(context: frank-small 2026-10-06 RCO-IQ3_XXS + draft-mtp cutover (commits da36fa3/bb3532d): live probes on the deployed two-slot shape; the pre-boot VRAM estimate was off by 6x on the draft ctx)_
|
||||
- HuggingFace repos on Xet storage (newer repos like ajgazin/Swift-1.5-Qwen3.8-27B-Uncensored-Dynamic-MTP-GGUF) expose NO LFS sha256 in the tree or blob APIs (the blob API returns a 404-ish "Sorry, we can't find the page" JSON for LFS files and resolve headers carry no X-Linked-Entity-Sha256). Verification for such repos must be: byte count against the Hub's declared size + local sha256sum after the first verified fetch, with the computed sha pinned in the download script (state in the script header that the sha was computed locally, not taken from the Hub). Proven on the 2026-10-06 frank-small retarget (both files matched Hub byte counts exactly; shas pinned in scripts/download-frank-small.sh).
|
||||
- radiance checkpoint reader FILE-SHADOW law: the reader opens every file the safetensors index names, in SORTED file-name order, and emplaces EVERY tensor each file physically holds into one table, first-wins (checkpoint.cpp add_shard, std::unordered_map::emplace never overwrites) — the index only enumerates files, it does NOT scope which tensors count. Consequence: a replacement file whose tensors share a name with tensors physically present in an earlier-sorting file is shadowed (its data never read; the plan reads the older file's tensor under that name). Fix: name the replacement so it sorts FIRST ("00-ple.merged.safetensors" beats "model-00001-of-00131.safetensors"), then patch the index value. Companion engine fact: the radiance engine's directory create is NOT recursive — "cannot create /data/kvcache/..." ENOENT left the SSD prefix tier off until the serve script mkdir -p'd the full path; pre-creating the host dir activated the tier on the running container without a repo deploy (volume exposes it immediately; only the boot-time check is one-shot). _(context: frank-primary radiance cutover, 2026-10-06 live bring-up; root-caused from /tmp/radiance-src checkpoint.cpp add_shard + shard_list after the plan-only run read a raw fp8 shard as the PLE codes)_
|
||||
- Fleet serve-script trap: podman run -d ... "$LOG" captures ONLY the container ID — the engine's stdout/stderr go to podman's internal store, so any runbook claiming "the boot ledger lives in $LOG" is reading a one-line hex file. Diagnosis signature: a ledger grep that returns empty, or a log file containing a single 64-hex string. Fix = run podman run FOREGROUND (Type=simple unit keeps the podman client as main process; podman run returns the container's exit code so Restart=always/ExecStop semantics are unchanged, and the client forwards SIGTERM). Before grepping any fleet service's log file, verify it is not this shape — frank-primary is suspect (its serve script uses podman run -d; its memory note claims logs land in /home/dank/logs/frank_primary.log, unverified against the -d trap as of 2026-10-06). _(context: frank-large tuning session 2026-10-06: the runbook's "engine stdout goes to /home/dank/logs/frank_large.log" had been false since the 2026-10-05 cutover — podman run -d ... "$LOG" captured only the container ID hash; the boot ledger existed only in podman logs frank-large-halogen. Fixed by foreground podman run (ba67ca5); AGENTS.md and the runbook now document it.)_
|
||||
- frank-large (halogen 0.1.4, Dankclaw iGPU) decode-at-depth laws, measured 2026-10-06 on real traffic + controlled A/B: (1) Decode collapses linearly with context: 30-50 t/s shallow → 5.5 @46k → 3.3-3.8 @80k; round-time slope ~8 µs/token ⇒ KV pool reads ~8 GB/s effective (with verify-position factor ~28 GB/s) vs weights streaming ~180-240 GB/s. Not an FA tile artifact — HALOGEN_FA_TQ=256 A/B'd same-prompt at 80k: decode ±0.3%, prefill −9.5% on this topology (the knob helps only the 64GB-carve crash topology of upstream issue #12). (2) The engine binary advertises 40+ UNDOCUMENTED HALOGEN_* kernel levers (FA_*/DN_*/MATMUL_*/ATTN_SD family) — enumerable read-only via grep -aoE "HALOGEN_[A-Z0-9_]+" /usr/local/bin/halogen inside the image with --entrypoint sh (no GPU touched; strings is NOT in the image, use grep -a). (3) Same-prompt depth A/B method: stop+mask prod, boot probe containers on 127.0.0.1:17770, prompt cache ON (8192 MB) so the second rep per arm skips the ~178s 80k prefill; force 128-token ESSAY instructions or greedy answers in 2 tokens and the decode measurement is void; read timings.predicted_per_second/draft_n_accepted. (4) Applied mitigation: harness roles slow+advisor (the deep-conversation roles) rerouted from frank-large to frank-primary in ~/.omp/agent/config.yml — frank-primary's fp8 KV is 7.32 KiB/token vs halogen's 64 (8.7x less KV read at depth). frank-primary's depth-decode rate is NOT yet measured on record — first deep session there should record timings. (5) Boot ledger WARNING "HALOGEN_CACHE_MB smaller than ONE full-context entry (18.4 GB at kMaxCtx=262144)" is a FALSE POSITIVE when slot_ctx native: entries size to their prompt (~70 KB/token); at slot 200,700 the deepest possible entry is ~14 GB, inside a 16384 MB pin. _(context: frank-large halogen tuning session 2026-10-06: prompt-cache pin (16384 MB) deployed and verified (16x warm prefill); decode-at-depth collapse investigated — FA_TQ A/B in a stop window rejected it (±0.3% decode, −9.5% prefill on Dankclaw); slow/advisor roles rerouted to frank-primary (config.yml, operator home). Commits ba67ca5, d39424f, fadad18.)_
|
||||
- radiance rad-convert from an HF safetensors checkpoint has FOUR non-obvious traps, each costing a full verify+declare cycle: (1) The reader's tensor table is FILE-DERIVED, not index-driven: checkpoint.cpp add_shard opens every index-named file in SORTED order and emplaces EVERY tensor each file physically holds, first-wins (std::unordered_map::emplace never overwrites). An HF checkpoint whose shards physically carry tensors under the SAME name a later replacement file uses will shadow it — e.g. the official Qwen3.8-Flash-Next-FP8 shards carry ...ple.ple_embedding.ngram_embedding.shard_0.weight ([2500012,160] fp8) which shadows a merged table of the identical name. Fix: name the replacement file so it sorts FIRST ("00-ple.merged.safetensors" beats "model-00001-of-00131.safetensors"). (2) scale_name() appends "_scale_inv" to the TENSOR NAME verbatim (checkpoint.cpp:521): a stacked tensor named ...experts.gate_up_proj (no .weight suffix) has its scale at ...experts.gate_up_proj_scale_inv — NOT ...gate_up_proj.weight_scale_inv. (3) Safetensors index.json/config.json in a staging dir are usually RELATIVE SYMLINKS to the raw tree; open(path,"w") writes THROUGH them and corrupts the raw checkpoint (verify then fails 143/144 on size). Always temp+os.replace for atomic swap — it replaces the symlink itself, leaving the target pristine. (4) numpy views over mmap: np.frombuffer keeps an exported pointer; mmap.close() raises "BufferError: cannot close exported pointers exist" (Python 3.14+) unless the view is del'd first. Also: rad-convert --plan-only prices the DATA planes; the written container carries a metadata tail on top (172.76 GiB plan - 177,109 MiB actual), so size floor gates must be set from the real file, not the plan. _(context: frank-primary radiance native-FP8 cutover, 2026-10-05/06; four convert blockers root-caused against /tmp/radiance-src (v1.0.7); all fixes landed in scripts/stack-qwen38-official-fp8.py + scripts/build-frank-primary-rad.sh)_
|
||||
- frank-large (halogen 0.1.4, Dankclaw iGPU) runtime facts, measured 2026-10-06 on real traffic + probes: (1) DECODE-AT-DEPTH COLLAPSE IS STRUCTURAL — 30-50 t/s shallow (1k, commit 4-6.3/round) → 5.5 @ 46k → 3.3-3.8 @ 80k; round-time slope ~8 µs/token of context = the KV path (16 full-attention layers, 64 KiB/token d2 KV) reads at ~8 GB/s effective while weights stream 180-240 GB/s; commit also falls to 2.5-3 at depth. No SUPPORTED flag touches decode (docs/FLAGS.md: W4A4/W4A4_EXCL are prefill-only) and 0.1.4 is the latest release; the publisher's tg bench NEVER measured depth (tools/halogen-bench.py tg rows run over the ten SHORT eval prompts), so README's "decode barely degrades at depth" is unexercised. (2) PROMPT CACHE was auto-OFF (engine floor = 2 × full-context entry = 36.9 GB vs 35.7 free); pinned HALOGEN_CACHE_MB=16384 (one full-slot entry ≈ 14 GB at ~70 KB/token); probe-proven cold 27.0 s → warm 1.7 s prefill at 17.3k (cache_n 16,384, prompt_n 946). Boot WARNING "smaller than ONE full-context entry (18.4 GB at kMaxCtx=262144)" is a FALSE POSITIVE vs our 200,700 slot — documented as expected in the runbook checklist. (3) SERVE LOG FIX: podman run -d LOG captures only the container ID — the engine ledger existed nowhere on disk; foreground podman run makes the ledger land in /home/dank/logs/frank_large.log with unchanged exit-code/ExecStop semantics; after a manual stop the unit reads "failed" (cosmetic). (4) UNDOCUMENTED LEVERS: the binary reads 40+ HALOGEN_* kernel-selection vars; enumerate with grep -aoE "HALOGEN_[A-Z0-9_]+" /usr/local/bin/halogen | sort -u (image has NO strings binary). Public issue #12 (64GB-carve box): default k_attn_fa2 crashes at PREFILL, HALOGEN_ATTN_FA=2 + HALOGEN_FA_TQ=256 restores full prefill speed — their decode was FINE, so FA_TQ's decode-at-depth relevance on Dankclaw (healthy prefill, so not in the crash class) is UNPROVEN; A/B window (default vs FA_TQ=256, 80k filler, 2×128-token decode, timings-read) was Frank-approved and i _(context: frank-large halogen tuning session 2026-10-06: cache pin shipped (ba67ca5, probe-verified 16x warm prefill), decode-at-depth diagnosed structural, FA_TQ A/B window approved by Frank and launched (arm 0 default vs arm 1 HALOGEN_FA_TQ=256); routing fallback mapped (slow/advisor → frank-primary). Commits: ba67ca5 (cache+foreground log), d39424f (docs+depth law).)_
|
||||
- podman log-capture trap (found 2026-10-06 on frank-large, likely affects frank-primary too): podman run -d ... "$LOG" 2&1 captures ONLY the container ID hash — the engine's stdout/stderr go to podman's internal log store, so the documented "boot ledger lives in /home/dank/logs/svc.log" is false and every runbook grep of that file reads a one-line hex string. Diagnosis: the log file contains exactly one 64-hex-char line. Fix that preserves systemd semantics (container exit code = service exit, ExecStop=podman stop still works, SIGTERM forwarded to the engine): run podman run FOREGROUND (drop -d, drop exec podman wait), keeping the "$LOG" redirect. Do not reach for --log-driver hacks (k8s-file JSON-wraps lines and races the shell redirect). Evidence + fix live in scripts/serve-frank-large.sh (commit ba67ca5); check frank-primary's serve script for the same gap before trusting its log file. _(context: frank-large tuning session 2026-10-05/06 (halogen on Dankclaw): runbook claimed engine ledger lands in /home/dank/logs/frank_large.log, but the file held one container-ID line; foreground-run fix shipped and verified (ledger grep now works).)_
|
||||
- Radiance (v1.0.7) convert/payload laws, proven on the frank-primary native-FP8 cutover 2026-10-05/06: (1) READER TABLE SHADOWING: Checkpoint::add_shard emplaces EVERY tensor each opened file physically holds, first-wins (checkpoint.cpp), and shard_list opens index-named files in SORTED name order — so a raw shard whose tensor name collides with a staging-file tensor shadows it if the raw file sorts first. Fix pattern: prefix the staging file so it sorts FIRST ("00-…"), letting its tensor win the emplace race. This bit when the plugin asked for …ple.ple_embedding.ngram_embedding.shard_0.weight and got the raw fp8 shard [2500012,160] instead of the merged bf16 table. (2) scale_name() = tensor name + "_scale_inv" VERBATIM (checkpoint.cpp:521) — stacked tensors (no .weight suffix) take …gate_up_proj_scale_inv, NOT …gate_up_proj.weight_scale_inv. (3) Staging symlinks raw files; ANY non-atomic rewrite of a staged json (open(path,"w")) writes THROUGH the symlink into the raw checkpoint — use temp + os.replace, or copy index/config into the stage instead of symlinking. Measured: raw index.json restamped with the stacked layout, raw verify then failed 143/144. (4) Python 3.14 BufferError "cannot close exported pointers exist": numpy frombuffer views keep mmap exports alive — del the view before mmap.close(). (5) Plan-only measures the .rad at the checkpoint's own fp8 mass (official Qwen FP8 131 shards = 172.76 GiB / 185.5 GB; the ~161 GiB estimate was wrong) — size any floor gate from the plan line, not estimates. (6) The plugin's bf16 branch (table_bf16 = plain-bf16 one-plane encoding) vs fp8 branch (codes + ONE-element scale plane) is decided at declare from the checkpoint form; a bf16 checkpoint + ONE rtn fixed-scale rule reproduces the publisher's dense→fp8 recipe path with exact code round-trip. (7) Debug hygiene: bg watchers that tail a sorted error list clip it — the tail of a sorted list is NOT the failure scope; fetch the full list before counting failures. All of t _(context: frank-primary radiance native-FP8 TP=2 cutover on dankserver; four failed convert attempts (BufferError crash, raw-index symlink corruption, stacked scale names, PLE table shadowing) each root-caused in /tmp/radiance-src before the clean plan-only at 172.76 GiB)_
|
||||
- Radiance rad-convert pipeline laws (proven live on dankserver 2026-10-05/06, frank-primary native-FP8 build; all three plan blockers root-caused from source, not guessed): (1) FILE-TABLE SHADOW: Checkpoint::add_shard (core/format/checkpoint.cpp) emplaces EVERY tensor each opened file physically holds into one name-tensor table, FIRST-WINS; shard_list opens files in SORTED name order. A merged/replacement tensor sharing a name with a tensor physically present in another index-named file is shadowed by whichever file sorts first. Symptom: declare sees the checkpoint form of a DIFFERENT tensor (e.g. merged PLE bf16 [320001536,160] vs "codes ... is [2500012,160]" = one raw fp8 shard inside model-000NN). Fix: name the new file so it sorts FIRST ("00-ple.merged.safetensors") — index values + script constants must follow. Corollary: the index's weight_map decides which files OPEN, not which tensors EXIST; 127 raw shard entries removed from the index still shadow via file contents. (2) SCALE-NAME LAW: checkpoint.cpp scale_name() = tensor name + "_scale_inv" VERBATIM. Stacked expert tensors have no .weight segment, so the scale plane is ...experts.gate_up_proj_scale_inv (NOT ...experts.gate_up_proj.weight_scale_inv). Wrong name = "FP8 with no scale plane beside it" on EVERY sliced expert (~100k errors). Also: a tail-N watcher artifact shows only the sorted TAIL of an error list — never size an error set from a tail (21 lines shown were the tail of ~100k). (3) NEVER write a staged file via open(path,"w") when staging uses symlinks: staging symlinks raw files into the stage dir, and the rewrite goes THROUGH the link into the raw tree (mtime proves it; raw verify then fails 143/144). Law: temp file + os.replace (replaces the link itself), and COPY (never symlink) any file the reshape step rewrites. Committed as fb1a4fd. (4) Python 3.14 mmap: np.frombuffer(mm,...) views keep the mmap exported; explicit mm.close() raises "BufferError: cannot close exported pointers exist". Law: d _(context: Session 2026-10-06: frank-primary radiance native-FP8 cutover on dankserver; three successive --plan-only failures (100,352 expert scale-name errors; symlink-through raw-index corruption found by the raw verify 143/144; PLE codes shadowed to one raw shard), each root-caused in /tmp/radiance-src source (checkpoint.cpp add_shard/shard_list/scale_name, rad_arch.h map_concat_shards, rad_block_ple.h fp)_
|
||||
- Two durable notes from the 2026-10-06 frank-small two-slot reshape: (1) For ANY llama.cpp what-if ("can we use X flag / drop a knob"), grep the SERVING BINARY's --help over ssh FIRST — verified live on Dankclaw's llama-master build before promising flags: the spec-type enum is draft-simple/draft-eagle3/draft-mtp/draft-dflash/draft-dspark/ngram-simple/ngram-map-k/ngram-map-k4v/ngram-mod (ngram family has its own --spec-ngram-mod-n-match etc.), and there is NO checkpoint knob anywhere — "limit checkpoints" in llama.cpp maps to dropping the draft context (draftless ngram types need no ctx_dft) and the internal recurrent-state rollback checkpoints carry no user flag. Answering from memory of other builds would have invented a knob. (2) omp bash-tool quirk: backgrounding a job with async:true requires timeout=null and name=null — passing a timeout alongside async errors out with "Service mode does not accept async or timeout"; the shape that works is {command, i, cwd, pty:false, async:true, timeout:null, name:null}, result then arrives as bg_N. _(context: 2026-10-06 frank-small two-slot no-spec reshape (commits 812d9eb, 5e16654): priced the what-if via the serving binary's --help over ssh before answering, and hit a bash-tool arg rejection when backgrounding the suite)_
|
||||
- Staged-tree rewrites MUST use temp+os.replace, never open(path, "w"): a staging tree that symlinks raw inputs lets a rewrite follow the link and corrupt the pristine source (measured 2026-10-05/06 on frank-primary radiance reshape: stack script rewrote stage's model.safetensors.index.json THROUGH the relative symlink, restamping the raw index with the stacked layout — 605 KB smaller, "ple.merged.safetensors" FileNotFoundError in the raw tree, raw verify failed 143/144). os.replace(tmp, path) swaps the symlink itself for a real file; the raw tree stays untouched. Corollaries: (1) a copy step beats a symlink for any staged file a later step rewrites — copy index.json/config.json, symlink only the big immutable shards; (2) recovery WITHOUT a 217 GB re-stack: safetensors headers can be patched in place — parse hlen u64 + JSON, rename keys, re-serialize, pad with b" " to the EXACT old hlen (json.loads tolerates trailing spaces; data_offsets unchanged), rewrite bytes 8..8+hlen; (3) disk-wall arithmetic: mv-to-trash is a rename and frees NO bytes, so a full re-stage only fits if free space ≥ old payload + new payload; 209 GB free vs ~217 GB stage meant the surgical header patch was the only path without a purge approval; (4) HF index.json sanity checks: total_size lives under metadata, not top-level (KeyError trap), and "listing says" size mismatches on ONE file with 143/144 passing = corruption/restamp, not a Hub revision. _(context: frank-primary radiance native-FP8 reshape, 2026-10-05/06: run 3 rewrote the staged index through a relative symlink into the raw checkpoint; overnight verify failed 143/144. Fixed in commits 272acb4/fb1a4fd (scripts/stack-qwen38-official-fp8.py temp+os.replace; scripts/build-frank-primary-rad.sh copies index.json instead of symlinking); recovered without re-stacking by patching 49 stacked headers )_
|
||||
- ComfyUI 0.37 /prompt API requires the node graph wrapped as {"prompt": {...}} — a bare graph body (the pre-0.3.x format) returns {"error": {"type": "no_prompt"}} while curl still reports full Content-Length sent. Verify renders via /history/prompt_id → status.status_str=="success", outputs in /home/dank/comfyui-data/output/. Also note: this repo's parallel sessions run broad git add commits — untracked new files can be swept into a concurrent session's commit (my ComfyUI workflow JSONs landed in frank-small commit baf33d0); after committing, verify git show --stat HEAD actually contains YOUR files before assuming. _(context: ComfyUI Qwen-Image workflow goal session 2026-10-05 (Dankclaw, port 8188, ComfyUI 0.37.0): two repair cycles to find the API wrap; parallel session baf33d0 swept the untracked JSONs.)_
|
||||
- Radiance rad-convert checkpoint laws (frank-primary native-FP8 build, 2026-10-05): (1) SCALE NAME LAW: checkpoint.cpp scale_name() appends "_scale_inv" to the tensor name VERBATIM — a STACKED expert tensor (…mlp.experts.gate_up_proj, no .weight suffix) has its scale plane at …mlp.experts.gate_up_proj_scale_inv, NOT …gate_up_proj.weight_scale_inv; the per-expert convention (experts.e.gate_proj.weight → .weight_scale_inv) does not carry over. Slice law: both codes and scale are sliced at the same sub-tensor index (checkpoint.cpp:579-589), so the stacked scale plane must be [E, r/128, c/128]. (2) Python 3.14 mmap law: np.frombuffer(mm,…) arrays hold exported buffer views — sm.close() raises BufferError: cannot close exported pointers exist unless the view is del'd before close; the loop's last view is always alive at close. (3) WATCHER LAW: tail -N of a convert log is NOT the error count — errors emit sorted lexicographically so the tail shows the sort's last names (.9/.80-.99), making a full 100k-error wall look like 21 strays; always grep -c '^E ' the whole log before classifying a failure. All three already fixed in scripts/stack-qwen38-official-fp8.py (commit 272acb4). _(context: frank-primary radiance native-FP8 cutover, 2026-10-05: first plan-only run failed with ~100k expert-name errors; root cause = stacked scale plane named with .weight; second failure = BufferError crash in PLE merge; misread tail of sorted error list as the full set before reading scale_name() in /tmp/radiance-src/core/format/checkpoint.cpp:579.)_
|
||||
- Two lessons from the 2026-10-05 Swift-1.5 frank-small cutover (shared checkout, live parallel writer): (1) A background suite run (bg_3) reported docs-seam test 135 failing while the LIVE parallel docs writer was editing the very index it tested — the writer's 23:25 commit fixed the seam between bg_3's snapshot and my recheck. Before chasing a seam failure that "shouldn't exist," stat the suspect files' mtimes against the suite start time and re-run the seam alone; a shared-checkout bg result is a snapshot of a moving tree, not ground truth. (2) PUT N: inserts after line N — if line N is a WRAPPED bullet/paragraph (continuation lines below), the insert lands mid-paragraph and the original continuation lines orphan below the new block. Anchor gap-inserts on the LAST line of a logical block, never its first, and re-read the whole region after any gap insert; this cost three repair edits on the runbook Trail section (bogus pasted fragment + orphaned continuation). _(context: 2026-10-05 Swift-1.5 IQ3_S frank-small cutover session; docs seam race with a live parallel session, and a PUT N gap-insert that landed mid-sentence inside a wrapped bullet)_
|
||||
- radiance/qwen4exp payload laws (2026-10-05, frank-primary native-FP8 cutover): (1) The shipped radiance qwen4exp_fp8 plugin maps ONLY the STACKED expert layout (mlp.experts.gate_up_proj [E,2F,H] + _scale_inv plane at the same index, gate rows first) while the official Qwen FP8 HF repo ships per-expert unfused tensors (experts.e.{gate,up,down}_proj.weight + weight_scale_inv each) — rad-convert refuses with 100,352 "needs checkpoint tensor ... not in stage" errors; 100,352 = 49 stacks (48 trunk + mtp.layers.0) x 512 experts x 4 tensors. Fix = staging byte-stack (pure concat, no requant). (2) The PLE table ships as 128 SCALELESS fp8 shards + ONE bf16 scalar scale (1.9931793212890625e-04); the plugin's fp8 branch declares the scale view as {1} — per-shard-blocked scales can never fit, and rtn on raw scaleless codes overflows (v/1.99e-4 e4m3's 448). Fix = staging decodes to one bf16 shard (exact), then ONE --quant rule 'blk.*.ple_ngram.weight=rtn:codes=fp8_e4m3,block=*x*,scale=bf16,rule=fixed,scale_value=the checkpoint's own scale' re-encodes at that same scale — provable exact round-trip (e4m3 grid 2^-3 vs bf16 rounding perturbation 2^-9), verified empirically for 12 representative codes + the 448 clamp. container PLE stays fp8 (47.7 GiB pinned, unchanged ledger). (3) The checkpoint name map gotchas: the shard regex captures pre WITHOUT the '.ple.' segment — scale and merged names must re-add it ({pre}.ple.ple_embedding.ngram_embedding.weight_scale). (4) TTM on the current dankserver kernel exposes /sys/module/ttm/parameters/pages_limit ONLY (no pages_max — the two-knob story is stale); sysfs raise to 52,428,800 pages = 200 GiB works at runtime; persistence needs ttm.pages_limit=52428800 on the kernel cmdline at next reboot; a serve-script preflight must SKIP absent knobs, not read them as 0. (5) Python 3.14 (current CachyOS) + numpy over mmap: np.frombuffer(mmap,...) arrays EXPORT the buffer — mmap.close() then raises 'BufferError: cannot close exported pointers exist' _(context: frank-primary radiance cutover session 2026-10-05: rad-convert plan-only triage (100,353 errors), staging reshape script written/debugged (name bugs, byte math, BufferError), one --quant PLE rule, TTM raise, commits 1919ad7+baf33d0 pushed)_
|
||||
- Radiance qwen4exp FP8 cutover reshape laws (2026-10-05, proven in source + on-host): (1) The shipped qwen4exp_fp8 plugin maps the STACKED expert layout ONLY (mlp.experts.gate_up_proj [E,2F,H] + mlp.experts.down_proj [E,H,F] with _weight_scale_inv planes at the same index = name+"_scale_inv"); the official Qwen3.8-Flash-Next-FP8 repo ships per-expert unfused tensors — the fix is staging-side byte-stacking (no requant; gate rows first inside each expert's slab), done by scripts/stack-qwen38-official-fp8.py with per-file sampled byte-identity asserts. (2) The PLE table's plugin fp8 branch wants codes beside a ONE-ELEMENT bf16 scale plane (decl_view {1} hardcode); a per-shard-blocked scale can NEVER fit it, and rtn on raw scaleless codes overflows (v/1.99e-4 e4m3 max 448) — the fix: staging decodes the 128 scaleless fp8 shards to ONE bf16 shard with the checkpoint's own scale (1.9931793212890625e-04), then ONE --quant rule blk.*.ple_ngram.weight=rtn:codes=fp8_e4m3,block=*x*,scale=bf16,rule=fixed,scale_value=s re-encodes at that same s — an exact round-trip (code grid 2^-3 vs bf16 perturbation 2^-9); the engine's bf16-vs-fp8 branch keys off the RESOLVED encoding, so the rule lands it in the fp8 branch. (3) TTM on current CachyOS kernels exposes ONLY /sys/module/ttm/parameters/pages_limit (NO pages_max); a preflight that reads an absent knob as 0 refuses a healthy cap — raise pages_limit (52,428,800 pages = 200 GiB done 2026-10-05; sysfs write doesn't survive reboot, cmdline ttm.pages_limit= is the persistence). (4) safetensors reshape scripts: compute index total_size from declared tensor bytes, never from a writer's return value (a header-only return silently undercounts ~116 GB). Repository of record: scripts/stack-qwen38-official-fp8.py docstring, scripts/build-frank-primary-rad.sh header, serverAppConfigs-docs/LLM/frank-primary-radiance-qwen38-native-fp8.md. _(context: frank-primary radiance cutover session 2026-10-05: convert blocked with 100,353 plan errors (100,352 expert-plane + 1 PLE); three source-reading sessions on /tmp/radiance-src (rad_block_moe_fp8.h, rad_block_ple.h, qwen4exp_fp8.cpp, TOOLS.md --quant grammar, lq_grid.cpp fixed rule) resolved it into two staging transforms + one recipe rule; round-trip exactness verified on-host for 12 representative)_
|
||||
- radiance qwen4exp_fp8 plugin vs official Qwen/Qwen3.8-Flash-Next-FP8 repo — LAYOUT LAWS (2026-10-05 cutover): the shipped plugin maps (1) STACKED experts — per MoE layer ONE tensor pair mlp.experts.gate_up_proj [E,2F,H] + mlp.experts.down_proj [E,H,F] with _scale_inv planes at the same dim-0 index, names under model.language_model.layers.N. and mtp.layers.0.; the official repo ships per-expert UNFUSED (experts.e.{gate,up,down}_proj.weight, 6 tensors/expert) → rad-convert --plan-only fails 100,352 errors (49×512×4); (2) PLE table fp8 branch wants codes + a ONE-ELEMENT bf16 scale plane (view {1}), but the repo ships 128 SCALELESS fp8 shards + one bf16 scalar (1.9931793212890625e-04, = the publisher recipe's own scale_value in TOOLS.md). FIX (scripts/stack-qwen38-official-fp8.py, wired into build-frank-primary-rad.sh): stage-time reshape, both transforms value-preserving — experts byte-stacked (fp8 codes moved, never decoded; gate rows first; fused scale = gate scale ++ up scale per expert), PLE decoded e4m3→bf16 with its own scale (LUT decode exact; RNE pack), then ONE --quant rule blk.*.ple_ngram.weight=rtn:codes=fp8_e4m3,block=*x*,scale=bf16,rule=fixed,scale_value=the checkpoint's own scale re-encodes at that scale = EXACT code round-trip (e4m3 grid 2^-3, bf16 rounding perturbation 2^-9 — cannot cross a midpoint; verified empirically incl. the 448 clamp). rtn on RAW scaleless codes would instead overflow (v/s ≈ ×5017 e4m3 448 max). Also: rtn re-encoding must never be applied to the scaleless form; a per-shard-blocked scale can NEVER fit the plugin's {1} view; rtn on scaleless-fp8 input destroys values. Pitfalls hit: name regex consumed the .ple. segment (scale + merged names must rebuild …layers.1.ple.ple_embedding.ngram_embedding.*); bf16 byte widths need explicit ×2 (element counts ≠ bytes); write_safetensors must return data bytes (header-only return mis-states logs and index total_size). TTM LAW (dankserver kernel): only /sys/module/ttm/parameters/pages_limit ex _(context: frank-primary radiance native-FP8 cutover session 2026-10-05: convert blocker triage (100,353 plan-only errors → 2-class fix), staging transform authored + round-trip verified on host, TTM cap raised pre-cutover, incumbent FreeToken stopped at Frank's order. Runbook + script docstring carry the full argument.)_
|
||||
- Converting the official Qwen FP8 checkpoint for the shipped radiance qwen4exp plugin (frank-primary cutover 2026-10-05): the plugin maps (a) STACKED experts — mlp.experts.gate_up_proj [E,2F,H] + down_proj [E,H,F] with _scale_inv planes at the same dim-0 index — so the official repo's per-expert unfused tensors must be BYTE-STACKED in staging (gate rows first inside each gate_up slab; pure byte concat, no requant); 100,352 plan errors = 49 stacks (48 trunk + mtp.layers.0) × 512 experts × 4 tensors (weight+scale each) — the arithmetic confirms which names are missing. (b) The PLE table's fp8 branch wants codes beside a ONE-ELEMENT bf16 scale plane (decl_view {1}); a per-shard-blocked scale can never fit that view AND rtn on the raw SCALELESS fp8 codes overflows (v/1.99e-4 ≫ e4m3's 448). The working transform: staging decodes the 128 scaleless shards with the checkpoint's OWN bf16 scale (0.00019931793212890625 — matches the publisher recipe's 1.99317932128906e-4 in TOOLS.md:141 exactly) into one bf16 shard_0 (config split_ngram_parts: 1), and the convert's single --quant rule blk.*.ple_ngram.weight=rtn:codes=fp8_e4m3,block=*x*,scale=bf16,rule=fixed,scale_value=s re-encodes at that same scale — an EXACT round-trip (nearest-code distance 2⁻³ vs bf16 rounding perturbation 2⁻⁹; empirically verified for 12 representative codes + the 448 clamp). Helper facts: e4m3fn decode is a 256-entry LUT (0x38=1.0, 0x7E=448, 0x7F=NaN only), f32→bf16 RNE = (bits+0x7FFF+((bits16)&1))16; np.argmin over a LUT with NaN entries returns the NaN index (mask with where(isfinite,…,inf) in tests). TTM on dankserver kernel 7.2.3: ONLY /sys/module/ttm/parameters/pages_limit exists (no pages_max — "both knobs" is stale); raised to 52,428,800 pages = 200 GiB pre-cutover (sysfs write, survives until reboot; persistence = ttm.pages_limit=52428800 kernel cmdline). Any cap function must SKIP an absent knob, not read 0. Also: dankserver host has numpy 2.3.5; the dev box does not. _(context: frank-primary radiance native-FP8 TP=2 cutover session, 2026-10-05; scripts/stack-qwen38-official-fp8.py, runbook serverAppConfigs-docs/LLM/frank-primary-radiance-qwen38-native-fp8.md)_
|
||||
- frank-primary radiance cutover, convert-blocker fix (2026-10-05): the official Qwen3.8-Flash-Next-FP8 checkpoint cannot rad-convert directly because the shipped qwen4exp plugin maps ONLY the stacked expert layout (mlp.experts.gate_up_proj [512,1280,2560] + _scale_inv at the same index; error-count arithmetic 49 stacks × 512 experts × 4 tensors = the 100,352 errors) and a PLE fp8 branch whose scale view is hard-coded {1}. Fix (all value-preserving, scripts/stack-qwen38-official-fp8.py + ONE --quant rule): (1) byte-stack per-expert gate/up/down + scales along dim 0 — no requantisation; (2) decode the 128 SCALELESS PLE fp8 shards with the checkpoint's OWN bf16 scale into one bf16 shard_0 (e4m3 values and the scale are exact in bf16), then blk.*.ple_ngram.weight=rtn:codes=fp8_e4m3,block=*x*,scale=bf16,rule=fixed,scale_value=ckpt scale re-encodes at that same scale — nearest-code distance 2^-3 vs the bf16 perturbation 2^-9, so every original code is restored exactly and the one-element scale plane lands as the plugin declares (rtn on the RAW scaleless codes would overflow: v/1.99e-4 exceeds e4m3's 448). Empirically verified on the host for representative codes incl. the 448 clamp. Safetensors writing: 8B LE u64 header length + JSON header, data_offsets relative to data start, pad header with spaces to 8-byte alignment; bf16 = RNE pack of f32 via (bits + 0x7FFF + ((bits16)&1)) 16. Dev box has NO numpy — test numpy-dependent host scripts on dankserver (numpy 2.3.5). Also: hashline PUT/CUT with remembered line numbers across MULTIPLE edits in one file silently corrupts (a mis-aimed PUT N.=M ate main-body lines and a later insert landed mid-function) — after any multi-edit sequence, py_compile BEFORE running; recover by re-reading the touched region and restoring eaten lines first. TTM on dankserver: /sys/module/ttm/parameters/pages_limit exists, NO pages_max; raised to 52,428,800 pages (200 GiB) pre-cutover; sysfs does not survive reboot (cmdline ttm.pages_limit=52428800 pe _(context: frank-primary radiance cutover 2026-10-05: stacking/PLE reshape + convert-blocker triage + TTM raise; runbook frank-primary-radiance-qwen38-native-fp8.md)_
|
||||
- TTM pinned-page cap on the CachyOS fleet hosts (verified 2026-10-05 on dankserver, kernel 7.2.3): only /sys/module/ttm/parameters/pages_limit exists — there is NO pages_max, so any preflight that mins over both knobs must SKIP an absent knob, not read it as 0 (a 0 makes a healthy cap refuse; fixed in scripts/serve-frank-primary.sh ttm_cap_mib). Pre-cutover cap was 32,252,917 pages = 123 GiB, BELOW frank-primary's 134 GiB pinned ledger. Raising is a runtime sysfs write (echo 52428800 | sudo tee ...pages_limit = 200 GiB) — allocates nothing, harmless, no GPU experiment rules apply; verified read-back. The write does NOT survive reboot: persist via kernel cmdline ttm.pages_limit=52428800 at the next scheduled reboot. Required pages = MiB × 256 (4 KiB pages); the serve gate is FRANK_PRIMARY_TTM_MIN_MIB=140000. _(context: radiance cutover pre-work for frank-primary (dankserver), 2026-10-05: TTM preflight gate design + live sysfs raise before the pinned-RAM engine boot.)_
|
||||
- Converting a scaleless-FP8 artifact into a container that hard-codes a one-element scale plane (radiance plugin law, e.g. rad_block_ple.h wants fp8 codes + a single bf16 scale): NEVER rtn-re-encode the raw codes at the fixed scale — v/1.99e-4 overflows e4m3's 448 max and the codes become garbage. The exact recipe: (1) staging DECODES the scaleless shards with the artifact's OWN bf16 scale into a bf16 table (e4m3 decode is exact; the product may round to bf16 by 2^-9), then (2) the convert's rtn rule re-encodes at that same scale — nearest-code distance is 2^-3 while the bf16 perturbation is 2^-9, so every original code is restored EXACTLY and the one-element scale plane is produced. Round-trip proven empirically for representative codes incl. the 448 clamp. Companion facts: a per-shard-blocked scale can NEVER fit a {1} decl_view (plugin hard-codes extents); safetensors bf16 = 2 B/element (byte-stacking code that forgets the ×2 produces ValueError/buffer-shortage crashes, and argmin over an e4m3 LUT must mask NaN or it returns the NaN entry 0x7F). Implemented in scripts/stack-qwen38-official-fp8.py + one --quant rule in scripts/build-frank-primary-rad.sh; full argument in the script docstring and serverAppConfigs-docs/LLM/frank-primary-radiance-qwen38-native-fp8.md. _(context: radiance native-FP8 TP=2 cutover for frank-primary (dankserver), 2026-10-05: convert of Qwen/Qwen3.8-Flash-Next-FP8 blocked by 100,353 plan errors (100,352 stacked-expert + 1 PLE scale-plane); derived the decode-reencode round-trip from rad_block_ple.h/rad_block_moe_fp8.h/lq_grid.cpp source tracing and verified it on-host before committing to a 102 GB staging write.)_
|
||||
- TWO lessons from the 2026-10-05 frank-large halogen cutover: (1) ROGUE/PARALLEL-SESSION RECOVERY RECIPE (this checkout is shared): a second omp session force-committed OVER this session's uncommitted work (serve script reverted to the pre-fix --ulimit nofile version, AGENTS.md lessons deleted, 413 restored over 400), then push created divergence mid-deploy. Recovery that worked: git log --graph --all to map the two lineages (rogue 3ee88d7/77decb5 Ornith commits + my be840b4/b0b840b above shared base bdbe814), git diff origin/main HEAD --stat to classify every diverging file as MINE (keep) vs ROGUE (drop), git rebase origin/main (clean replay of my 2 commits), then re-apply the specific reverted hunks. The seam is the tripwire that catches this class: tests/frank_small_unit.bats failed 7 tests because the rogue swap commit (Ornith payload) never touched the deploy handler + seam twins — stale seams DETECT half-finished swaps. Also: my own edit-tool tags went stale after the clobber — trust grep, not remembered line numbers, and re-read every region before editing. (2) halogen (frank-large) contract facts verified LIVE 2026-10-05 (engine 0.1.4, health reports version.api==version.engine==0.1.4): boot ledger prints "kv pool: 1 slot(s) x 200700 tokens = 13.15 GB" + dtype census (bf16 2.96 GB, q4c 9.93 GB, fp8r 10.63 GB, i4l 12.35 GB — p1w4d-d2). Prompt cache AUTO SIZES ITSELF OFF on this box ("35.7 GB free after mapped checkpoint ... less than the 36.9 GB floor (2 x 18.4 GB full-context entry)") — expected, not a fault. Drafter verified from the RESPONSE timings object (llama-server shape): draft_n/draft_n_accepted counts at 39-42 t/s greedy — DFlash2 active on the response, not just /health's drafter_default. Tool round-trip REAL: get_weather - {"city":"Seattle"} as an OpenAI tool_calls array (wire format qwen-xml inside the engine). Contract: over-cap max_tokens AND prompt-over-slot AND response_format are ALL 400s naming the limit (README-verified; my earlier 413 cla _(context: frank-large halogen cutover session 2026-10-05; discovered when a parallel omp session force-committed over uncommitted work mid-deploy)_
|
||||
- Parallel-session docs writer vs the deploy tree gate: deploy.sh refuses on ANY uncommitted change INCLUDING untracked files (whole-tree guard, scripts/deploy.sh ~line 117), and live docs writers (Frank's Obsidian vault work or another agent) constantly produce untracked serverAppConfigs-docs/LLM/*.md files. Stashing a LIVE writer's strays is a data-loss race: 2026-10-05 a stash -u of 6 stray docs was followed within seconds by 3 MORE untracked files appearing — the writer was active; had the stash held, their in-flight files would have vanished from under them. Reversed with an immediate git stash pop (safe: stash snapshots, nothing lost). Working protocol: (1) check writer liveness first — mtimes within ~10 min of now means LIVE, do not stash; (2) the labeled-stash convention ("parallel-session topic WIP - stashed for reason") applies only to quiet strays; (3) if the writer is live, the deploy is genuinely blocked on Frank's coordination — say exactly which files block the whole-tree gate and ask him to commit/close the writer; do NOT commit another session's WIP yourself, do NOT stash twice in a race, and do NOT bypass the gate. _(context: 2026-10-05 frank-small Ornith deploy: 364/366 tests green, swap pushed, download verified — deploy blocked solely by a live docs writer's strays at the whole-tree gate; one stash/pop round-trip proved the race is real before it cost anything.)_
|
||||
- Recipe: read a remote HF GGUF's metadata without downloading it — HTTP range request + struct parse. curl -fsSL -r 0-8388607 -o head.bin https://huggingface.co/repo/resolve/main/file.gguf, then parse: magic 4B, version u32, tensor_count u64, kv_count u64, then KV pairs. Two parser traps cost three failed attempts (2026-10-05, Ornith-1.5-9B sizing): (1) initialize offset to 4 AFTER the magic assert — treating offset 0 as the version field reads the magic itself as u32 (0x46554747 = "GGUF") and every downstream field shifts; (2) string arrays must be fully consumed (n × (u64 len + bytes)) — early-returning "str[n]" without consuming leaves offsets inside array data and later keys misread as unknown types. The tokenizer vocab array (~150k strings) sits BETWEEN the general.* keys and the architecture keys (qwen35.block_count etc.), so 64 KiB is NOT enough — fetch ~8 MiB. Yields arch name, block_count, head counts, key/value lengths, full_attention_interval (→ KV rate = full_attn_layers × kv_heads × klen × 2 B × 2), nextn_predict_layers, expert_count (absent = dense), native context — enough to size VRAM before committing a payload swap. Note: an existing managed skill gguf-loader-compatibility-check may be the right home for this; verify its content before minting a duplicate. _(context: 2026-10-05 frank-small→Ornith-1.5-9B swap: needed Ornith's architecture (9 full-attn layers, 36 KiB/token f16 KV) to size the shape; the range-read settled it in two cheap requests against the live Hub file, no download required.)_
|
||||
- Frank's standing working rules for skill management (established 2026-10-05): (1) Do NOT use subagents for reading/analysis/distillation work — he explicitly ordered solo work ("Without subagents… read through each skill", later "NO. SUBAGENTS."); parallel fanout is not authorized by default even when tools suggest it. (2) The knowledgebase (serverAppConfigs-docs) is the system of record for all fleet knowledge — managed skills are NOT durable homes for lessons; when skills duplicate or extend runbooks, distill their content into vault pages and delete the skills, staging deletions in ~/.AGENTTRASH first. His default is "no reason to keep these skills". _(context: 2026-10-05 managed-skills audit session: Frank twice ordered "Without subagents" and then "NO. SUBAGENTS." when the agent planned task-fanout for reading 173 skills and distilling them; and he ruled that none of the fleet skills are worth keeping as skills — their content belongs in serverAppConfigs-docs.)_
|
||||
- serverAppConfig multi-session triage rule (proven 2026-10-05, three events in one day): when ./tests/run.sh goes red and the failing seams point at files you did not touch, FIRST split the breakage by git state: (a) COMMITTED-but-broken states from a parallel session (swap commits that forgot their seam/deployer twins) → repair them yourself in clearly-labeled follow-up commits; that is what kept the suite green twice. (b) UNCOMMITTED working-tree state (M/untracked files owned by a LIVE concurrent session — check git status --porcelain and untracked ?? entries against HEAD, which may not have moved) → do NOT edit those files; the other session is mid-flight and will reconcile env+seam+docs in its own commit; editing its dirty tree risks colliding with an active writer. Verify your own pushed state instead (rev-list HEAD...origin/main, targeted seams on your files) and report the named dirty files as the red-suite owner. Signature of case (b): HEAD identical across two suite runs yet different failures, because bats seams read the working tree, not the commit. _(context: 2026-10-05 goal session: frankbook e2e split (9c50821/2930431/1c725ea/5573ca1) interleaved with a live parallel session committing Ornith-swap breakage and later holding the tree dirty (M frank-small.env, M frank_small_unit.bats, ?? GGUF Compatibility.md).)_
|
||||
- Wrapper-script law (serverAppConfigs tests/e2e/run.sh): when converting a pure passthrough exec tool "$@" into "default argument set + passthrough", the default set must be APPENDED to "$@", not substituted for it — exec npx playwright test "${PROJECT_ARGS[@]}" silently DROPS every other arg (--list, --grep, file filters), so run.sh --list ran the entire 34-test suite instead of listing. The regression is only visible by diffing observed behavior against requested semantics; the seam (bash -n) cannot see it. Correct shape: build PROJECT_ARGS only when no user --project is present, then exec npx playwright test "$@" "${PROJECT_ARGS[@]}". Caught 2026-10-05 because the --list proof printed execution checkmarks instead of a listing; verify arg-forwarding wrappers by exercising EVERY documented passthrough arg, not just the happy path. _(context: 2026-10-05 frankbook e2e split; regression introduced by my own default-project-set rewrite and caught during verification (commit 1c725ea).)_
|
||||
- frank-small seam congruence rule (serverAppConfigs repo): ANY change to scripts/frank-small.env has TWO consumers that must flip in the same change: (1) scripts/deploy-remote.sh's LLM_SMALL_MODEL_FILE/LLM_SMALL_MMPROJ defaults (the payload-existence gate — test 196 greps for the exact env path inside the deployer default), and (2) tests/frank_small_unit.bats (TARGET_STUB/MM_STUB fixture names, the MODEL_FILE env assertion, and the KV-contract test pinning --cache-type-k/--cache-type-v values from the env). On 2026-10-05 two consecutive commits broke it: the Ornith-1.5-9B Q8_0 payload swap (77decb5) left both consumers on MiMo-Q6K paths, and the KV reprice (1862dc3, f16 V→q8_0 V) left the KV test asserting the old unquantized contract — its commit message even claimed the seam was 'already retargeted in 8f53df7' when 8f53df7 covered only payload paths. Symptom each time: full ./tests/run.sh red in frank_small_unit.bats only ('missing /tmp/.../mmproj-...' or argv-value mismatch). Fix = grep the env for the new values and update all three spots. _(context: 2026-10-05 goal session (frankbook e2e split) blocked twice by frank-small seam staleness from concurrent sessions; fixed via 8f53df7-equivalent edits and 5573ca1.)_
|
||||
@@ -0,0 +1 @@
|
||||
Current fleet is stable but marginal on `dankserver` RAM (4600 MT/s). `frank-large` (Halogen) and `frank-primary` (VLLM) are the main inference providers. Watch for `max_tokens` 400 errors on large prompts and orphaned `llama-server` processes after benching. `frank-medium` requires 16384 MiB `CACHE_RAM` for hybrid DeltaNet states. Quadlet units ensure boot persistence. Use `ak shell` for Authentik user management. Distill skills to vaults to reduce context bloat (~38k tokens overhead).
|
||||
File diff suppressed because it is too large
Load Diff
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a09648-8eb8-7093-876a-116b11d4c5fb
|
||||
updated_at: 1789246675
|
||||
|
||||
Deployed marreta on dankclaw via podman behind the Anubis+Authentik guard chain, sharing gluetun egress with searxng. Resolved login-wall caused by missing Application link in Authentik ProxyProvider (null flows crashing Rust outpost). Implemented provider-first route flip, netns-aware probing, and e2e regression test.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a09686-789e-7042-8774-55507faa78b2
|
||||
updated_at: 1789232480
|
||||
|
||||
frank-large is a rootless Quadlet service on dankclaw. Logs retrieved via podman logs frank-large or journalctl --user -u frank-large. Key constraint: avoid naming TEMP/HOST/GPU_MAX_HEAP_SIZE env vars. Weights pinned (~68 GiB).
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a09835-91cf-72e7-8ece-e5316d4de48b
|
||||
updated_at: 1789286336
|
||||
|
||||
Expanded frank-large KV pool to 524288 positions (~14.4 GiB) and reduced slots from 4 to 3. frank-large defaults to greedy; frank-small defaults to sampling (1.0 temp, top_k 20, top_p 0.95). preserve_thinking is per-request only. 285/286 tests pass.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a09e67-630f-74ed-a910-e997e01fbe54
|
||||
updated_at: 1789368883
|
||||
|
||||
Halogen-flash-server docs say multi-minute full-CPU stalls with no disk activity/output under co-tenancy/pinned trunk are documented behavior, not crashes; do not restart. Last resort: HALOGEN_FLASH_PIN_TRUNK=0 returns memory but costs several times decode speed. Related durable constraints: first request after restart may be silent ~45-60s and self-recovers; frank-large maxTokens must be <=65536 and prompt+max_tokens must fit the KV pool or requests queue then 503.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0a7ad-42e2-739d-bc1c-dc405ca02cbd
|
||||
updated_at: 1789574671
|
||||
|
||||
Configured dual AMD R9700 GPUs on dankserver for ROCmFPX llama.cpp. Discovered RCCL fails on chipset-attached PCIe slot (broadcast/allreduce/send/recv all fail), killing both TP and PP. Solution: $30 M.2→PCIe x4 riser to move card #2 to CPU-attached M.2_2 slot. Documented in serverAppConfigs-docs/LLM/Dual-R9700-Cooperation.md (commit cc29875). Host went down during probing; needs power cycle.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0a87f-6b17-7492-995c-8c3e7b37e241
|
||||
updated_at: 1789552363
|
||||
|
||||
Deployed mradermacher/Qwen3.8-Flash-Next-Uncensored-i1-GGUF (i1-Q4_K_M) on dankclaw via ik_llama.cpp with CUDA (3090) + Vulkan (Strix Halo gfx1151). Shipped at ~3.8-4.0 t/s decode, ~192 t/s prefill. 30 t/s decode proven unreachable: 100 graph splits/token cause PCIe x4 launch/sync bottleneck; CUDA graph capture aborts on qwen4exp hc ops (ggml_repeat_4d is capture-illegal). Mainline PR #28901 is the only path to 30 t/s class. Root systemd unit on :7770, contextWindow 262144.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0a902-7f58-77f4-9392-764f6cfbd59e
|
||||
updated_at: 1789544414
|
||||
|
||||
Rewrote serverAppConfigs AGENTS.md (734→412 lines, commit f227786): collapsed swap-history into active-topology table + runbook pointer table, added 'Key Lessons Learned' section with durable ROCm/llama.cpp/quadlet pitfalls, condensed halogen section to invariants only. Also diagnosed and fixed missing omp skills: pi-era skills in ~/.pi/agent/skills/ and ~/.agents/skills/ are invisible to omp (which only scans ~/.omp/agent/skills/); restored 9 of 13 from mattpocock/skills and vercel-labs/skills repos.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0ac6b-61a9-7505-b00a-52dd4cbb4ff3
|
||||
updated_at: 1789599423
|
||||
|
||||
Testing thread. Story response surfaced durable technical context: rootless Podman `restart: unless-stopped` does NOT survive explicit shutdown (only crashes), so bare containers vanish on reboot. Repo `serverAppConfig` mandates Quadlet `.container` files for boot persistence. Historical incidents: Aug 26 Caddy lost after reboot; Sep 11 five observability services found unitless.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0ac72-2620-750f-82dd-79c84e00e98d
|
||||
updated_at: 1789631572
|
||||
|
||||
User asked about M.2 slots on dankclaw, clarified they meant the Qwen Flash pipeline-parallel deployment. Dankclaw is an AMD Strix Halo APU (gfx1151) with an M.2 dock carrying an RTX 3090 at PCIe x4. The M.2 slot docs in the repo are split: dankserver (B650-E, M.2_2 free for R9700 riser) vs dankclaw (M.2 dock for 3090). The Qwen Flash pipeline-parallel M.2 question was aborted before being answered.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0aeb5-e0ba-742d-86da-df30e865107d
|
||||
updated_at: 1790026831
|
||||
|
||||
Multi-phase llama.cpp private fork project planned: clone mainline, merge charlie12345 ROCmFP4 quantization, merge ik_llama.cpp improvements, TDD throughout, optimize for gfx1201 (dankserver 2x R9700 layer-split) and gfx1151 (dankclaw iGPU). Session interrupted by connection failure to frank-medium (dankserver) and context overflow on frank-large (dankclaw, 144k prompt + 57k max_tokens > 200k context).
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0b09a-aeae-75f5-9522-c5a3f1d97ac7
|
||||
updated_at: 1789710503
|
||||
|
||||
Verified fleet health across dankclaw/dankserver/frankbook-pro. Deployed frank-medium 35:30 tensor split (VRAM headroom on bottleneck card tripled to 3.00 GiB; layer lever exhausted at 36:29). Diagnosed and confirmed fix for frank-large max_tokens 400 dead zone (maxTokens 57344 in models.yml closes the gap; old OMP sessions must be restarted to pick up config). 326 seams green.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0b23a-7a27-7792-b8da-97d7421b99c3
|
||||
updated_at: 1789756944
|
||||
|
||||
Diagnosing Plex container read-permission failure on /mnt/entropy at dankserver. Constraint: cannot modify /mnt/entropy permissions; fix must be container-side via GroupAdd. Session interrupted before final fix was applied and verified.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0b30f-1e89-763e-b65d-48d87e6ee683
|
||||
updated_at: 1789756946
|
||||
|
||||
frank-medium (192.168.0.235:7770) found stopped after an incomplete bench protocol run: the final `systemctl stop` at 04:25:59 was never followed by unmask+start+health-verify. Restored safely (GPU idle check → start → health 200 → /proc/MainPID/cmdline geometry verify → behavioral completion). Durable lesson: the restore half of the bench protocol is not optional; a session that benches owes the fleet a health 200 before it ends. Consider making the bench harness self-restoring via trap EXIT.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0b5e1-46b7-7387-a64d-f6f69f79be10
|
||||
updated_at: 1789847739
|
||||
|
||||
Fleet model-swap workflow on Dankclaw: switch frank-small to vLLM with gemma-4-12B QAT w4a16 + MTP speculative decoding (assistant model kunhunjon/gemma-4-12B-it-qat-assistant-w4a16-ct). Post-cutover fleet state: frank-small (:7771) Gemma4 Q4_K_M, frank-large (:7770) CIRU Orca, frank-medium orcarouter Q8_0+DFlash2. 3090 RGB controlled via EVGA iCX3 EEPROM zone-mode registers (m0-m3=0x00 for OFF, persists across reboot without Linux daemon).
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0b5f3-22ec-7692-a98c-8290e5cc3476
|
||||
updated_at: 1789762341
|
||||
|
||||
frank-medium outage on 2026-09-18 (second of the day) caused by an orphaned merge-build llama-server (PID 2519965, PPID 1, bound 127.0.0.1:17770) holding 15.5+17.1 GiB across both R9700s after a successful load test that was never cleaned up. Unit was in clean-stopped state (inactive/enabled/unmasked/NRestarts=0). Diagnosed via rocm-smi showing busy GPUs despite inactive unit; proved abandonment via frozen /proc/pid/stat, zero established conns, loopback-only bind, 3h stale log. Fixed by killing stray and restarting frank-medium (503→200 in ~20s). Logged as commit cfabe07.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0b693-7972-7104-bb4b-2bbec7f42acc
|
||||
updated_at: 1789773472
|
||||
|
||||
Doubled CACHE_RAM 8192→16384 MiB for frank-medium (llama.cpp, Qwen3.8-27B hybrid DeltaNet on 2x R9700). The old cap refused real states (137 refusals in 2 days, up to 11.4 GiB) because a full slot state is ~12.8 GiB at 196608 ctx (68 KiB/token from 16 full-attn layers + MTP draft KV). With --ctx-checkpoints 0, a refused state means full re-prefill next turn — the RAM cache is the only cross-slot state carrier for hybrid DeltaNet models.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0c09e-b740-7547-8216-e31373fa0316
|
||||
updated_at: 1789968641
|
||||
|
||||
Two topics: (1) vllm-radiance TP=2 is impossible on Dankserver's 2x R9700 (gfx1201) because the chipset lacks PCIe atomics (Gen4 x1, AtomicOps Routing-), which RCCL collectives require; the workaround was two single-GPU replicas behind a round-robin proxy, later replaced by llama.cpp DFlash2. (2) Updated stale comments in ~/.omp/agent/models.yml after MTP→DFlash2 drafter swap: model is now Qwen3.8-27B-Uncensored-FP8 (30.9GB/7 shards), accept 4.24/7, KV pool shrank from 943,581 to 384,238 tokens (3.59→1.47 concurrent) due to DFlash2's 2.1GiB drafter cost.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0c510-1716-724c-b99b-b1ee4df08178
|
||||
updated_at: 1790020691
|
||||
|
||||
Calibre Content Server deployed on Dankserver (192.168.0.235:8084) with four critical traps discovered: LS image 9.15.0 is desktop not headless, internal port 8080 is nginx so use 8090, default auth is Digest not Basic, and echo-piped passwords include trailing newline. Calibre-Web 0.6.27 confirmed to have no generic OIDC; guard chain is the only SSO path.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0c51e-53af-724a-9f4e-364858819889
|
||||
updated_at: 1790063680
|
||||
|
||||
frank-medium (vLLM/radiance on dankserver) already supports --chat-template via $CHAT_TEMPLATE in radiance-entrypoint.sh line 253. The Qwen-Sharp chat template is archived at scripts/.AGENTTRASH/chat-template.jinja.ik-llama-qwen-sharp. Applying it to frank-medium requires pointing CHAT_TEMPLATE to that file (or a copy in a stable location). The frank-large swap (Kairic Edge → new model + DFlash2 drafter) was completed in commits 7689092 + 600812d.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0c58d-9751-7094-bb98-3d7f54669e56
|
||||
updated_at: 1790024021
|
||||
|
||||
Explored homepage service architecture (services.json + app.js category-based rendering) and the three-script deployment pipeline (deploy.sh → deploy-plan.py → deploy-remote.sh) with path-aware service scoping, per-service state store for crash recovery, and a Playwright e2e gate against the live fleet.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0c5c1-0f4c-72bc-81b8-8b05f92d0ef1
|
||||
updated_at: 1790063676
|
||||
|
||||
Explored homepage architecture (shared/homepage/ with services.json, app.js, style.css) for section-based reorganization; discovered Open WebUI image generation API endpoint at /api/images/generate in the open-webui podman container on dankclaw.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0c5e9-dee3-734a-946c-3cf809f9d7fd
|
||||
updated_at: 1790027980
|
||||
|
||||
OMP session persistence architecture: JSONL files at ~/.omp/agent/sessions/<project-slug>/<timestamp>_<id>.jsonl; blob externalization (strings>500K truncated, image data URLs always content-addressed as blob:sha256:<hash>, other base64 images at 1024 chars); storage backends (File/Memory/Indexed); discovery reads 4KiB prefix + 32KiB tail; HistoryStorage is separate SQLite at ~/.omp/agent/history.db for prompt recall.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0c6e4-d2b6-7078-a0c9-f492a78605a2
|
||||
updated_at: 1790063108
|
||||
|
||||
frank-medium swapped to ParoQuant MXFP6 (Swift Qwen3.8-27B) + Qwen-Sharp v22.5.0 template. GPU_UTIL=0.99 proven infeasible on 32GB R9700 (vLLM pre-weight-load check fails: 31.54 GiB required > 31.23 GiB free); ceiling is 0.98. MXFP6 weights 25.16 GiB (vs MXFP4 19.37 GiB) → KV pool 425,047 tokens (1.62x). DFlash2-FP8 drafter survives Swift finetune (acceptance 85.7%/58.3%).
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0c81c-56dc-7359-ac24-6da36332ad84
|
||||
updated_at: 1790068527
|
||||
|
||||
Exhaustive fact-check and OKF compliance pass on serverAppConfigs-docs vault (136 pages). Key corrections: ComfyUI 0.37.0 has native Qwen-Image 2.1 (custom GGUF patches may be redundant on rebuild); OpenWebUI imagegen endpoint is POST /api/v1/images/generations; CVE-2025-7404 is Calibre-Web not Calibre; GTT is 114 GiB not 96; deployment is Quadlet/deploy-plan.py based. Three research briefs added (local TTS/voice cloning, AI job search tools, popular self-hosted AI apps Jun-Sep 2026) plus HTML report. All 336 tests pass, 0 unresolved wikilinks.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0d06e-c8b5-757b-a093-2081af1db7c6
|
||||
updated_at: 1790503506
|
||||
|
||||
A 3000-word narrative depicting the invention of the 'Soviet burger'—a kotleta (meat patty) nestled in a pampushka bun. It explores the transition from a disorganized cafeteria meal to a standardized GOST product, highlighting the tension between bureaucratic perfection (the standard) and the variable 'life' of the kitchen.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0d66d-f9e2-7681-a5e3-e77b896dcc3f
|
||||
updated_at: 1790407605
|
||||
|
||||
Flipped homepage CSS stagger to da[REDACTED] (role/inputs dim, spec/use bright) and removed the unused --input variable. Verified deployment on www.franksdank.cloud. Confirmed Authentik + OpenWebUI integration feasibility via OAuth2/OIDC.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0df97-1e50-77d7-9084-92708bfd1152
|
||||
updated_at: 1790458743
|
||||
|
||||
Identified persistent connectivity issues with dankserver (frank-medium) and established the /guided-goal framework for structured objective setting.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0dfa8-7193-7235-980d-69e138218fcc
|
||||
updated_at: 1790474002
|
||||
|
||||
Deploying GLM-5-Next (GLM-5.3-Flash) 321.3B MoE model on dankserver. Key features include vision tower support, MTP draft mode, and a specific requirement for `-fa off` and `-ub 128` to handle MLA and long-context collapse. Since the Q6_K weights (~264GB) exceed the 246GB RAM, it uses `mmap` with `-ot exps=CPU` to offload experts to host RAM.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0e091-9152-75fc-b81b-8e311474eea8
|
||||
updated_at: 1790624016
|
||||
|
||||
Transitioned frank-primary to Qwen3.8-Flash-Next-GGUF at Q_8. Configured with Q_8 drafter, vision enabled, and layered speculative decoding using mtp and ngram-mod.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0e144-473f-71a5-96f8-0316232910aa
|
||||
updated_at: 1790503503
|
||||
|
||||
Established foundational engineering laws for the Frank fleet: -ot (tensor overrides) follows first-match-wins logic; -ot exps=CPU reduces PCIe overhead; spec-decoding limits (one context-backed drafter + multiple ngrams); ubatch optimum is 512; and GGUF extents must be derived from header offsets to avoid 'gap' buckets.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0e983-6c65-77e6-813a-c655e46ce06f
|
||||
updated_at: 1790628098
|
||||
|
||||
Deployed prefix caching and integrated autoresearch branch updates (PLE timeout 1800s, bench matrix runbook, index/doc-link fixes) for the frank-primary model on R9V hardware.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0ea08-9e09-7608-8003-81ab447e507e
|
||||
updated_at: 1790641878
|
||||
|
||||
Purged all documentation mentions of 'Dankware' (retired workstation) from both `~/.omp` and `serverAppConfigs`. Functional artifacts `packages/dankware.pkgs` and `scripts/install.sh` were preserved as the name is load-bearing for host lookups.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0eb77-a0b2-7748-8c3c-c2eee29d34d4
|
||||
updated_at: 1790724034
|
||||
|
||||
Replace the current llama.cpp-based runtime of frank-large with the gufo inference engine.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0f190-490b-765e-b62d-21e28c3c0ec0
|
||||
updated_at: 1790760416
|
||||
|
||||
Deep research on DeepSeek 4.1 Flash quantizations for the current hardware (2x R9700 gfx1201 + 246 GiB DDR5 RAM). Identified `vcruz305 Q2_K` as the optimal fit for the backbone (~191 GiB) and `GSQ-RCO 3.0-bit` as the high-quality alternative. Key bottleneck is RAM bandwidth (42 GB/s), limiting decode to ~6-11 tok/s. Next step: Build native HIP version from the fork.
|
||||
+4
@@ -0,0 +1,4 @@
|
||||
thread_id: 01a0fb3e-f1d3-7330-bd66-2b997c1a0264
|
||||
updated_at: 1790938297
|
||||
|
||||
Reconfigured frank-small to llama.cpp MiMo-V2.6-Distill-Qwen-9B-GGUF:Q5_K_M with 3 contexts (180,224 tok/slot) + vision on RTX 3090; frank-large to Qwen3.8-27B-ROCmFP4-STRIX-MTP-GGUF on gfx1151 iGPU. Key: n_ctx_slot=CTX_SIZE/PARALLEL formula, vision bills ~1.4 GiB, legacy images field ignored, compaction overflow risk.
|
||||
Some files were not shown because too many files have changed in this diff Show More
Reference in New Issue
Block a user