grenade

grenade pushed to main at helexa/helexa

2026-05-28 08:06:41 +00:00

0121a1930f feat(helexa-acp): inject and parse Qwen3 Hermes tool format

grenade pushed to main at helexa/helexa

2026-05-28 07:38:14 +00:00

13f4c36aeb chore(helexa-acp): log outgoing chat-completion body at debug level

grenade pushed to main at helexa/helexa

2026-05-28 07:33:20 +00:00

4a51a54554 fix(helexa-acp): describe Stage 3 tools in the default system prompt

grenade pushed to main at helexa/helexa

2026-05-28 07:01:41 +00:00

0609f1ac5d feat(helexa-acp): add tools, session modes, and permission gating

grenade pushed to main at helexa/helexa

2026-05-28 06:46:35 +00:00

96fc379893 feat(helexa-acp): wire ACP agent loop for text-only conversations

grenade pushed to main at helexa/helexa

2026-05-28 05:13:59 +00:00

e267f583e1 chore(neuron): rustfmt drift in is_device_fault test

e23d5011d0 feat(helexa-acp): scaffold ACP bridge with provider trait + OpenAI chat

Compare 2 commits »

grenade pushed to main at helexa/helexa

2026-05-27 15:57:51 +00:00

249b2e5c98 fix(neuron): only poison the model on actual device faults

c59da83636 fix(neuron): serialise single-GPU inference per loaded model

Compare 2 commits »

grenade opened issue helexa/helexa#2

2026-05-27 15:30:16 +00:00

Research: Prototype a quantization kernel in pure Rust via cuda-oxide

grenade pushed to main at helexa/helexa

2026-05-27 15:02:55 +00:00

f05882369d fix(neuron): don't poison the model on tokio JoinError panics

bd04d7f580 fix(neuron): stream tokens via DecodeStream to avoid UTF-8 panic

Compare 2 commits »

grenade pushed to main at helexa/helexa

2026-05-27 13:49:57 +00:00

1e13889392 feat(neuron): chunked prefill + VRAM/prompt-length pre-flight checks

grenade opened issue helexa/helexa#1

2026-05-27 10:31:42 +00:00

Reduce TP=2 Q6K cold-load time for Qwen3.6-27B (~5 min today)

grenade pushed to main at helexa/helexa

2026-05-27 10:13:18 +00:00

6e1c1dd0fc ci: retry clippy + test up to 3 times on spurious sccache failures

grenade pushed to main at helexa/helexa

2026-05-27 09:47:34 +00:00

35876954cd chore(neuron): default tracing filter to info (was info,neuron=debug)

grenade pushed to main at helexa/helexa

2026-05-27 09:36:25 +00:00

740299bd9d chore(neuron/beast): switch default-model quant from q5k to q6k

cdf0f4e66d fix(neuron): trim cudarc mempool after clear_kv_cache to release VRAM

Compare 2 commits »

grenade pushed to main at helexa/helexa

2026-05-27 08:15:48 +00:00

c4954e0eed docs: per-device worker thread architecture (phase 5 of refactor)

grenade pushed to main at helexa/helexa

2026-05-27 07:24:42 +00:00

b4f3576d82 refactor(neuron): phase 4 — model loads move onto the device worker

grenade pushed to main at helexa/helexa

2026-05-27 07:16:07 +00:00

76ab24d98c refactor(neuron): phase 3 — TP forward + NCCL state move onto device worker

grenade pushed to main at helexa/helexa

2026-05-27 06:55:11 +00:00

b179204fd3 refactor(neuron): phase 2 — single-GPU forward + clear_kv route through device worker

grenade pushed to main at helexa/helexa

2026-05-27 06:40:38 +00:00

081b532387 refactor(neuron): phase 1 — per-device worker thread, VRAM queries route through it

grenade pushed to main at helexa/helexa

2026-05-27 06:05:00 +00:00

7c19da9361 feat(neuron): construction-complete vram/config dump + logits health + per-step vram