chore(memory): snapshot auto-learned memory and renamed workstation skills
--home-blank-.omp--: MEMORY.md +140 / learned.md / memory_summary.md / raw_memories.md +1002 rewritten by auto-learn after the model rename, frank-medium split-balance and bottleneck studies. Skills renamed to single-topic buckets: dotfile-git-workflow -> ghostty-text-aa + niri-display-rotation + noctalia-theme, handy-service-mgmt -> handy-flm-inference, inference-runtime -> llama-cpp-parallelism. --home-blank-Projects-serverAppConfig--: 38 new rollout summaries, raw_memories.md/MEMORY.md refreshed, skills re-bucketed (bench-protocol, llm-fleet-model-swap, rootless-quadlet-log-retrieval, orchestration-quadlet). Per-worktree bucket (wt-20261006-122204) is now gitignored, not committed.
This commit is contained in:
1 parent
a34322aeb0
commit
5b08896e4e
72 files changed
+1461
-175
No files matched your search
+1
@@ -0,0 +1 @@
|
||||
SSH to `dankclaw`, run `ak shell`. Create User: `User.objects.create(username=..., email=..., is_active=True)`. Set password: `user.set_password("..."); user.save()`. Assign groups: `Group.objects.get(name="...").users.add(user)`. Verify via `authenticate(user, password)`.
|
||||
@@ -1 +0,0 @@
|
||||
The standard 'bench stop-and-mask' protocol. A session that benches owes the fleet a health 200 before it ends. Safe order: 1. Confirm GPUs idle. 2. systemctl start. 3. Health 200. 4. Verify geometry (/proc/MainPID/cmdline). 5. Behavioral proof (completion).
|
||||
-1
@@ -1 +0,0 @@
|
||||
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.
|
||||
+1
@@ -0,0 +1 @@
|
||||
Target: Strix Halo iGPU (3090 occupied). Quant: FP8 or Q6. Display: Ghostty `\e_G` base64 PNG. Harness read tool image behavior: `read <png>` → inline block (vision models), `read <png>?q=<question>` → text answer.
|
||||
+1
@@ -0,0 +1 @@
|
||||
175 managed-skills descriptions caused ~38k token overhead. Distill to vault reference pages (git commit). Move dirs to `~/.AGENTTRASH/`. Purge after explicit user approval (`purge!`).
|
||||
+1
@@ -0,0 +1 @@
|
||||
Stop-and-mask protocol must include restore half. Orphaned `llama-server` (PID 2519965, PPID 1) held GPUs after bench. Diag: `rocm-smi`, `ps -eo`, `cat /proc/<pid>/stat`. Fix: `kill <pid>`, `systemctl start`, verify geometry.
|
||||
+1
@@ -0,0 +1 @@
|
||||
4x DIMMs at 4800 MT/s caused silent corruption at ~40k tokens. Fix: Drop to 4600 MT/s. Consumer non-ECC RAM raises no MCE for silent flips. Short in-band checks miss intermittent corruption. Software reverts tested under same marginal hardware are confounded.
|
||||
+1
@@ -0,0 +1 @@
|
||||
Core stall: With `HALOGEN_FLASH_PIN_TRUNK=1`, co-tenant can stall engine for minutes. Last resort: `HALOGEN_FLASH_PIN_TRUNK=0`. First request after restart may be silent ~45-60s. `frank-large` `maxTokens` must be <=65536. Prompt + max_tokens must fit KV pool or queue ~3600s then 503s.
|
||||
+1
-1
@@ -1 +1 @@
|
||||
Workflow for switching a model on a host: 1. Read HF model card (runtime pin, constraints). 2. Check host for existing files. 3. Verify artifact integrity (size + sha256sum). 4. Deploy via git flow (commit -> push -> deploy.sh). 5. Live verification + port-holder health-check trap.
|
||||
Read HF model card first. Check host storage (`ls /home/dank/models/`, `df -h`). Verify artifact integrity (LFS oid = sha256). Deploy via git flow: commit → push → `deploy.sh <host> --service <name>` → health check → e2e.
|
||||
-1
@@ -1 +0,0 @@
|
||||
Contextualize model settings: max_model_len, KV pool, drafter acceptance, and weights.
|
||||
+1
@@ -0,0 +1 @@
|
||||
Standard deploy script with tree-clean gate.
|
||||
+1
@@ -0,0 +1 @@
|
||||
Rootless Podman `restart: unless-stopped` only survives crashes. Bare containers vanish on reboot. Repo `serverAppConfig` mandates Quadlet `.container` files. Historical: Caddy lost after reboot (Aug 26).
|
||||
-1
@@ -1 +0,0 @@
|
||||
Specific log retrieval for rootless podman units: 1. podman logs <unit> for full output. 2. podman logs <unit> 2>&1 | grep -E '^(dmalloc|kv pool):' for filtered metrics. 3. journalctl --user -u <unit> for systemd integration.
|
||||
-1
@@ -1 +0,0 @@
|
||||
podman logs frank_large 2>&1 | grep -E '^(dmalloc|kv pool):'
|
||||
Reference in new issue
Block a user