-
04a8fb7806
frank-default-small: PARALLEL=3 160k, KV K q8_0 / V q5_0 + Hadamard
Frank
2026-08-26 05:43:35 -07:00
-
8c41c821c4
frank-default-small: Q8_0 model, 160k single slot, q6_0 KV + Hadamard
Frank
2026-08-26 05:01:58 -07:00
-
23e905c712
frank-default-small: pull expert blocks 28,29 back to GPU
Frank
2026-08-26 04:05:24 -07:00
-
3f3b1bdb27
frank-default-small: render Qwen-Sharp via --jinja; pull more experts to GPU
Frank
2026-08-26 03:46:54 -07:00
-
4144f0f0e8
frank-default-small: disable speculative decoding to keep 3 slots
Frank
2026-08-26 03:38:25 -07:00
-
0e5c37f2f8
frank-default-small: use q8_0 (not q8_KV) KV cache on ik_llama
Frank
2026-08-26 03:23:18 -07:00
-
e2085f1f7d
frank-default-small: use split-mode none on single-GPU dankserver
Frank
2026-08-26 03:06:30 -07:00
-
b8de25d70b
frank-default-small: run on ik_llama.cpp with q8_KV + Hadamard cache
Frank
2026-08-26 02:59:34 -07:00
-
c8d7ed65c6
fix: use tls_insecure_skip_verify token (this Caddy build has no tls{} transport sub-block)
Frank
2026-08-26 02:16:42 -07:00
-
aa473aeddb
fix: proxy Plex through a TLS upstream (PMS drops plain HTTP off-loopback)
Frank
2026-08-26 01:24:04 -07:00
-
daeeb0d198
test: pin TLS upstream transport for Plex vhosts
Frank
2026-08-26 01:24:04 -07:00
-
67483982c3
Pin dankware resolver to Pi-hole on both default-route links
Frank
2026-08-26 01:21:36 -07:00
-
0150333a6a
serve-frank-large: restore Qwen-Sharp chat template + reformat for readability
Frank
2026-08-26 00:38:39 -07:00
-
8d8a5f7842
test-changed: map reorganized dankclaw quadlet subtrees
Frank
2026-08-25 22:59:54 -07:00
-
ab0ef3f4b7
Add change log metadata
Frank
2026-08-25 22:08:34 -07:00
-
786ef4a5cf
Preserve retired proxy state during migration
Frank
2026-08-25 22:02:14 -07:00
-
29af0c751f
Make canonical Caddy recreate safely
Frank
2026-08-25 22:01:12 -07:00
-
8110884e8e
Document canonical Caddy migration
Frank
2026-08-25 21:59:59 -07:00
-
d52bcc15f8
Migrate Gitea proxy to canonical Caddy
Frank
2026-08-25 21:56:04 -07:00
-
35e3e1a636
Set equal batch and ubatch at 8k context
Frank
2026-08-25 21:04:00 -07:00
-
d5b5b6a914
Cover independent observability planning
Frank
2026-08-25 20:36:04 -07:00
-
bc718f04f4
Fix scoped deployment state and config ownership
feat/service-scoped-deploy
Frank
2026-08-25 20:24:47 -07:00
-
3a56a07883
Guard target-owned deployment handlers
Frank
2026-08-25 20:10:16 -07:00
-
a07b8c9306
Deploy services independently by changed paths
Frank
2026-08-25 20:08:52 -07:00
-
1ba47e4d22
Deploy services independently by changed paths
Frank
2026-08-25 20:08:03 -07:00
-
2277a2e9ac
docs(kv): add bit-density KV cache quant table across 3 runtimes
Frank
2026-08-25 20:03:13 -07:00
-
4308f62004
Normalize Hermes local provider names
Frank
2026-08-25 19:42:46 -07:00
-
8fb5bd7a5a
Remove unrelated Hermes reasoning overrides
Frank
2026-08-25 19:33:56 -07:00
-
5766ff7dab
Set ubatch sweep to requested 32k matrix
Frank
2026-08-25 19:12:42 -07:00
-
c02d67513d
Update benchmark documentation after harness replacement
Frank
2026-08-25 19:08:42 -07:00
-
863897f92a
Clarify llama-bench harness scope
Frank
2026-08-25 19:07:46 -07:00
-
dc69e2b4b4
Support fixed ubatch batch sweep at 16K context
Frank
2026-08-25 18:54:09 -07:00
-
dfdd4bb6a8
Set Hermes local model reasoning levels
Frank
2026-08-25 18:48:04 -07:00
-
4e8ea24b2f
Record complete 32K ubatch benchmark findings
Frank
2026-08-25 18:40:32 -07:00
-
929796352a
Research llama.cpp runtimes + SGLang; fix AGENTS.md OKF spec link
Frank
2026-08-25 18:23:19 -07:00
-
2693259227
configure anonymous SearXNG provider defaults
deploy/searxng-provider-policy
Frank
2026-08-25 17:46:09 -07:00
-
e6670f914f
docs: record live SearXNG provider baseline
Frank
2026-08-25 17:30:42 -07:00
-
a05584b155
tune SearXNG VPN request timeouts and retries
Frank
2026-08-25 17:26:19 -07:00
-
e1f05bd3c2
Configure Hermes access to small LLM
Frank
2026-08-25 17:55:52 -07:00
-
96f5ae6fe1
Raise ubatch benchmark context to 32K
Frank
2026-08-25 17:44:45 -07:00
-
4cb67d7b33
Set ubatch sweep context and batch to 16K
Frank
2026-08-25 17:32:12 -07:00
-
babc823f11
Avoid shell variable collision in bench parser
Frank
2026-08-25 17:26:32 -07:00
-
56ee7cca45
Pass run number into ubatch benchmark parser
Frank
2026-08-25 17:12:51 -07:00
-
c072c75fa9
Make ubatch benchmark cleanup failure safe
Frank
2026-08-25 17:11:42 -07:00
-
119ca2493f
feat: add tmux session snapshot and restore utility
Frank
2026-08-25 17:04:15 -07:00
-
a8e0b88d99
Parse current llama timing output in ubatch bench
Frank
2026-08-25 16:57:02 -07:00
-
835996a67b
Keep deployment moving when Caddy restart is blocked
Frank
2026-08-25 16:55:10 -07:00
-
eae5bc2565
Continue LLM deployment when compose restart fails
Frank
2026-08-25 16:54:18 -07:00
-
c7a72b465e
Fix privileged shared config deployment
Frank
2026-08-25 16:53:13 -07:00
-
d6ea71b9a4
Configure DeepSeek reasoning and large-model ubatch sweep
Frank
2026-08-25 16:52:26 -07:00
-
8c8c3f7cd0
Enable focused tests as repository pre-commit hook
Frank
2026-08-25 16:18:01 -07:00
-
63ff1b6b43
Add focused changed-path test wrapper
Frank
2026-08-25 16:05:59 -07:00
-
7a0feb39aa
Add CLI helper for entropy volume
Frank
2026-08-25 15:38:57 -07:00
-
6ee195dd8f
Restore historical Q6 large model runtime
Frank
2026-08-25 14:23:47 -07:00
-
ddfc64013f
Fix Q6 AGENT projector path
Frank
2026-08-25 13:58:50 -07:00
-
86d5f38142
Restore Q6 AGENT large model
Frank
2026-08-25 13:36:07 -07:00
-
205b11aae4
Clarify historical ROCmFPX payload sizes
Frank
2026-08-25 13:28:38 -07:00
-
6c9e71efd2
Document Q8 AGENT set_rows serving blocker
Frank
2026-08-25 12:16:47 -07:00
-
54b2d84613
Upgrade frank-default-large to kingjones Q8_0 AGENT at full 262144 context
frank-default-large-q8agent-262144
Frank
2026-08-25 07:12:02 -07:00
-
24efa98d56
docs: add SearXNG providers guide for unblocking Google behind Gluetun VPN
Frank
2026-08-25 06:22:57 -07:00
-
b457457bfe
Switch large service to Kairic Edge v1.2
Frank
2026-08-25 04:41:04 -07:00
-
f86ed70c71
Use conservative four-token MTP drafts
Frank
2026-08-25 04:20:15 -07:00
-
3cf6353050
Reduce large model context to 64K
Frank
2026-08-25 04:13:00 -07:00
-
b0eb571321
Reduce large model context to 128K
Frank
2026-08-25 04:09:12 -07:00
-
ee3c3de750
Reduce frank-default-large context for KV headroom
Frank
2026-08-25 04:05:33 -07:00
-
7ca9dc248e
frank-default-large: swap to caf-eemin Qwen3.8-27B ROCmI4 embedded-MTP
eemin-rocmi4-mtp-large
Frank
2026-08-25 03:33:29 -07:00
-
7a8500f6fa
Tune frank-default-large params
frank-large-tune-ctx-batch-thinking
Frank
2026-08-25 02:30:25 -07:00
-
cd14d7bf68
Use ngram-mod and prioritize attention experts
tune-frank-small-context-163840
Frank
2026-08-25 01:40:06 -07:00
-
37fb420fcd
Record no-spec ubatch benchmark
Frank
2026-08-25 01:29:49 -07:00
-
f13124f977
Select hybrid expert placement from benchmark
Frank
2026-08-25 01:08:00 -07:00
-
339325719e
Reduce small model context and rebalance GPU layers
Frank
2026-08-25 00:59:41 -07:00
-
7a17984dbc
Tune small model expert placement
Frank
2026-08-25 00:31:05 -07:00
-
af6ffeeb64
Switch frank-default-small to Ornith Q6_K
Frank
2026-08-25 00:29:14 -07:00
-
3cc99d81cc
Authentik SSO + homepage naked domain
Frank
2026-08-24 23:48:27 -07:00
-
7ba1acdbc5
frank-default-small: lower context by 32k per stream for cache headroom
Frank
2026-08-24 21:50:54 -07:00
-
4462aa992c
Swap LLM fleet to Ornith CRACK (small) and kingjones ROCmFP4 STRIX MTP (large); apply Qwen-Sharp chat template
Frank
2026-08-24 21:35:11 -07:00
-
189934af15
Caddyfile: remove basicauth from chat.franksdank.cloud (OpenWebUI has its own auth)
Frank
2026-08-24 18:15:45 -07:00
-
35c3fc9ff5
Pi-hole records: mount addn-hosts into FTL hostsdir, not dnsmasq.d
Frank
2026-08-24 14:26:58 -07:00
-
0f59316be9
docs: Pi-hole KB page + dnsmasq retirement across the vault
Frank
2026-08-24 13:54:35 -07:00
-
bb96fc55f7
Pi-hole swap: quadlet + env + records + Caddy vhost (retires dnsmasq)
Frank
2026-08-24 13:54:35 -07:00
-
cf3da02af7
Add weekend report + doc-audit HTML (report/doc-audit-2026-08-23.html)
Frank
2026-08-24 01:22:43 -07:00
-
90c4ac69a7
Doc audit: fix all broken links + correct stale facts across the KB
Frank
2026-08-24 01:10:44 -07:00
-
ac7c8ee442
Caddyfile: point the shared family login at dankuser too
Frank
2026-08-23 22:26:46 -07:00
-
c58dab0a89
Homepage: give the user page its own login (dankuser)
Frank
2026-08-23 21:08:21 -07:00
-
677046768b
docs: record the deployed homepage + the Caddyfile handle design
Frank
2026-08-23 20:07:52 -07:00
-
5e6ba222cb
compose: point caddy's /srv/homepage at the host dir, not the repo
Frank
2026-08-23 19:18:41 -07:00
-
c7ced90aca
Caddyfile: serve the homepage via handle blocks (fix the /admin/ 401)
Frank
2026-08-23 18:33:54 -07:00
-
25af3d9fdf
Caddyfile: fix the /admin/ rewrite (matcher follows the directive)
Frank
2026-08-23 17:42:44 -07:00
-
e7eaba4457
Homepage (TDD slice 7): fix the /admin/ path + status.json perms
Frank
2026-08-23 16:49:33 -07:00
-
b485a27784
deploy: run the /srv/homepage steps under sudo
Frank
2026-08-23 15:44:46 -07:00
-
f400be76c6
deploy: make the Caddy hash grep defensive under set -e
Frank
2026-08-23 15:25:53 -07:00
-
cc1f421d27
Homepage: record the completed build in the KB
Frank
2026-08-23 13:06:58 -07:00
-
2e6c44f3ba
Homepage (TDD slice 6): the git-driven deploy
Frank
2026-08-23 12:59:56 -07:00
-
13ea2d7d3e
Homepage (TDD slice 5): the Caddyfile auth split
Frank
2026-08-23 12:32:18 -07:00
-
01c42202b8
Homepage (TDD slice 4): the pages, theme, and toggle
Frank
2026-08-23 12:26:31 -07:00
-
ce55d1dc3a
Homepage (TDD slice 3): the real services.json + schema tests
Frank
2026-08-23 12:11:21 -07:00
-
fcb6593afd
Homepage (TDD slice 2): the http probe method
Frank
2026-08-23 11:36:56 -07:00
-
05257df8e4
Homepage (TDD slice 1): the status checker core
Frank
2026-08-23 10:42:54 -07:00
-
e430b6dbad
Homepage: record all 38 settled decisions (grilling complete)
Frank
2026-08-23 10:31:15 -07:00
-
cdae928f37
Homepage: settle design (grilling Q1-Q5), log open questions Q6-Q43, progress report
Frank
2026-08-23 02:59:48 -07:00