187c717383
fix(pi extension): simplify regex for replacing 'pi' with 'claude code'
2026-05-11 09:06:30 -04:00
352e99c732
feat(llama-swap): add gemma-4-26b-vision model config
2026-05-10 16:59:35 -04:00
6fff658f9d
feat(llama-swap): add --default-chat-template-kwargs to vLLM 3090 configs
...
Sync all three Qwen3.6 27B vLLM configs (tools-text, long-text,
long-vision) with club-3090 83bf73d. Adds disable-thinking flag
and introduces upstream hash tracking comments for future syncs.
Update update-vllm-3090-configs skill to use hash-based skip logic.
2026-05-10 16:57:17 -04:00
885fe8517d
chore: remove fish
2026-05-10 09:27:48 -04:00
2133786985
chore: update flake.lock inputs
2026-05-10 08:36:23 -04:00
b41e9f2a84
docs(pi): add agent knowledge capture guidance
2026-05-09 10:18:13 -04:00
b25a933dd0
docs(pi): tighten agent guidance
2026-05-09 10:16:34 -04:00
37b0fae7e2
fix(llama-swap): sync qwen vllm 3090 configs
2026-05-09 10:16:32 -04:00
02410568dc
docs(skills): add vllm 3090 update workflow
2026-05-09 10:16:29 -04:00
2f29008a2a
chore: remove Handy speech-to-text dependency
2026-05-09 09:07:36 -04:00
87cdabbef4
feat(home): add Handy speech-to-text module
2026-05-07 22:36:30 -04:00
18a4e46bcb
chore(lin-va-thinkpad): enable nix-ld
2026-05-07 20:37:56 -04:00
f3cc67b17d
chore(llama-swap): tune presence penalty to 1.5 and remove repeat penalty
2026-05-07 20:37:54 -04:00
d142c5ff7e
build(pi-coding-agent): bump to 0.73.1 and migrate to earendil-works org
2026-05-07 20:37:53 -04:00
fea5cc887d
feat(llama-swap): add Qwen3.6-27B MTP thinking model and bump llama-cpp to b9045
...
Add qwen3.6-27b-mtp-thinking model config with 150K context, MTP
speculative decoding, and thinking mode support. Bump llama-cpp
from b9009 to b9045 and apply MTP patch from upstream PR #22673 .
2026-05-06 12:49:49 -04:00
f6f17831bf
feat(lin-va-terminal): enable nix-ld
2026-05-05 23:50:48 -04:00
5ded87b24a
chore(packages): bump pi-coding-agent to 0.73.0
2026-05-05 19:14:32 -04:00
d3ccbda958
refactor(llama-swap): update Genesis to v7.69 and upgrade vLLM nightly image
...
- Bump Genesis pin from 2db18df to 7b9fd319
- Upgrade vllm/vllm-openai nightly from 7a1eb8ac to 01d4d1ad
- Remove standalone boot-time patches now folded into Genesis (patch_tolist_cudagraph,
patch_inputs_embeds_optional, patch_workspace_lock_disable, patch_pn25_genesis_register_fix,
patch_pn30_dst_shaped_temp_fix, patch_pr40798_workspace)
- Reorganize environment variables across all vLLM compose configs
- Add new Genesis optimizations: P100, P101, P103, P15B, P38B, PN59 streaming GDN
2026-05-05 15:50:37 -04:00
3bccdc6382
chore(llama-swap): disable GENESIS_ENABLE_P85
2026-05-04 12:53:11 -04:00
0e3658615a
fix(home/pi): use config instead of osConfig for sops checks, cleanup formatting
2026-05-04 12:20:59 -04:00
3095515963
feat(conduit): add package and home module
2026-05-04 00:00:06 -04:00
4701a97a91
feat(pi): add statusline package
2026-05-03 11:41:36 -04:00
6e20baf883
feat(pi): merge sops-managed auth keys
2026-05-02 22:44:03 -04:00
f19894b45f
feat(lin-va-mbp-personal): enable nix-ld
2026-05-02 21:38:12 -04:00
d72b563276
build(glimpse): update package source
2026-05-02 20:34:51 -04:00
d1f17a18b4
docs(web-glimpse): simplify and streamline skill documentation
...
Condense the web-glimpse SKILL.md from verbose multi-section format to a
compact quick-reference style. Key changes:
- Consolidate usage patterns into a single quick reference block
- Replace separate sections per command with a concise command table
- Simplify workflow guidance and error handling into scannable tables
- Update timeout values from milliseconds to seconds
- Document new --no-reader and --format options
- Remove redundant answering guidelines
2026-05-02 20:32:36 -04:00
193b431681
chore(packages): bump llama-cpp to b9009
2026-05-02 17:19:19 -04:00
3117a3569f
chore(packages): bump pi-coding-agent to 0.72.1
2026-05-02 17:17:17 -04:00
3cc4649979
feat(lin-va-terminal): add sops secret for rke2 kubeconfig
2026-05-02 17:04:52 -04:00
7c1519881a
refactor(llama-swap): generate sops secrets from apiKeys list
2026-05-02 15:48:15 -04:00
f00edb620c
feat(lin-va-desktop): add systemd services for NVIDIA power limit
2026-05-02 14:00:49 -04:00
8d45977154
chore(llama-swap): update Genesis to v7.69, add cliff 2 optimizations
...
- Bump Genesis pin from fc89395 to 2db18df (v7.69)
- Add PN32 GDN chunked prefill and PN34 workspace lock relax env vars
- Replace patch_workspace_lock_disable with patch_inputs_embeds_optional
- Remove setup-time PN25/PN30 patches (folded into v7.69 natively)
- Switch patch base URL to v7.69-cliff2-test branch
- Lower GPU memory utilization to 0.93 for long-text variant
- Remove python3 from preflight check prerequisites
- Add printing service to lin-va-thinkpad
2026-05-02 13:48:41 -04:00
40114f438f
feat(llama-swap): sync vLLM configs from club-3090, add evan API key
...
Sync all three vLLM model configs from club-3090 master (ae4846f).
Update to Genesis v7.65 full PROD env set with new patches.
Update docker image to nightly-7a1eb8ac. Add torch_compile and
triton cache dirs. Add agent setup guide (AGENTS.md).
Add 'evan' API key to llama-swap sops secrets.
2026-05-02 08:27:47 -04:00
ba30222962
feat(pi): add skills and improve AGENTS.md reading guidelines
...
- Add 'create-skill' skill for scaffolding new skill directories
- Add 'planning' skill for structured implementation workflows
- Add search-then-read pattern guidance to AGENTS.md
2026-05-01 23:26:53 -04:00
e4d40d89d9
feat: add api keys to llama-swap
2026-05-01 22:12:51 -04:00
43a1d66e6b
add 9b vision
2026-05-01 21:51:06 -04:00
1283b7cdef
add fim 4b + 9b
2026-05-01 21:42:50 -04:00
09fdff4908
refactor(llama-swap): reorganize models by GPU hardware section
2026-05-01 21:08:53 -04:00
88308602c8
feat(llama-swap): add concurrent model matrix for CUDA0/CUDA1
...
Allow one CUDA0 and one CUDA1 model to run simultaneously. Dual-GPU
models (using -ts splits) are excluded from the matrix so they evict
everything when loaded. vLLM docker models get evict_cost=50 to
discourage eviction due to slow cold starts.
2026-05-01 16:50:28 -04:00
1812d2ea03
feat(pi): add vision model support
...
Extract a shared hasType helper for model filtering and add
vision (text + image) input capability to compatible models.
Also tag two llama-swap models with the vision type.
2026-05-01 15:03:26 -04:00
ab63211a75
wip
2026-05-01 14:36:36 -04:00
561f10d2a7
fix: timing & vllm
2026-05-01 13:09:28 -04:00
a3b2efa5bb
feat: vllm timings patch
2026-05-01 10:57:59 -04:00
74ff71803b
feat: vllm yay
2026-05-01 10:38:43 -04:00
75eba8703f
add: vllm base 3.6 27b
2026-04-30 21:47:29 -04:00
3d55b6e675
chore: add lfs
2026-04-30 20:14:41 -04:00
990b6a4392
feat: vllm
2026-04-30 20:04:58 -04:00
bcba8f6b60
feat(address-gh-review): add thread resolution with reactions and comments
2026-04-30 14:20:01 -04:00
93e2247a30
chore(nixos/llama-swap): remove synthetic peer and tune local model args
2026-04-30 11:43:04 -04:00
31363f5f8d
docs(pi): add _scratch guidance for ephemeral artifacts
2026-04-30 08:32:03 -04:00
