zhongqin
|
47354ca846
|
feat(agent): MCP client (stdio + streamable-HTTP JSON mode) with per-server timeout & response cap
|
2026-09-04 19:04:47 +08:00 |
|
zhongqin
|
9910956023
|
feat(agent): memory tiers — pinned protection + selective consolidation to Core
|
2026-09-04 16:37:17 +08:00 |
|
zhongqin
|
f1432fb957
|
feat(agent): context windowing — model-window budget, whole-turn LLM fold, canonical context:summary row, manual compact, reactive overflow fallback, expose cache_read/write usage (OpenAI/DeepSeek) + mini-epoch rebuild attribution; add cache & epoch smoke scripts
|
2026-09-04 14:38:30 +08:00 |
|
zhongqin
|
a4b83e0dc2
|
feat(agent): memory tiers — heuristic importance + zeroclaw budget eviction, per-turn hygiene. transcript context window — fold oldest turns to LLM summary (meta.ctx, default off)
|
2026-09-04 11:21:28 +08:00 |
|
zhongqin
|
39b8d139ee
|
feat(agent): gated run_shell with deny-list policy + sandboxed run_code polish , managed file tools over VirtualFs (tenant workspace)
|
2026-09-04 09:46:51 +08:00 |
|
zhongqin
|
d4da1084cb
|
feat(agent): skills pipeline — import CLI, PATCH agent, skill__tool composed, search_posts, turn observability & broadcast
|
2026-09-04 07:49:26 +08:00 |
|
zhongqin
|
abcf299d50
|
feat: add skills system
|
2026-09-04 06:19:24 +08:00 |
|