Query the witness-anchored ruLake cache — search, verify, explain, or refresh. Returns ranked results plus a decision_trace block with cost, latency, witness match, and substrates used.
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
Wraps the rulake_query MCP tool with sensible defaults so a fresh
operator can issue their first query without thinking about the
intent / risk / freshness / target axes.
search | verify | explain | refresh (default: search)low | medium | high (default: low){
"result": [ /* ranked retrieval */ ],
"decision_trace": {
"chosen_path": "deterministic-retrieval-path-v0.1",
"intent": "search",
"freshness": { "budget_ms": 5000, "actual_ms": 1031 },
"cache": { "hit": true, "hit_ratio_session": 0.87 },
"substrates_used": ["gcs-backend", "rvdna-backend"],
"kernel": { "id": "cpu-naive", "deterministic": true },
"witness": { "match": true },
"cost": { "compute_kernel": 0.0, "backend_fetch": 0.0, "cache_hit_discount": -1.0 },
"latency": { "total_ms": 1.02, "cache_ms": 0.4, "witness_ms": 0.6 },
"refusals": []
}
}
The decision_trace block is the new contract from ADR-009 — it lets a calling agent negotiate cost vs trust vs latency without reaching around the abstraction. Note: the trace shape ships in v2.4 — until then, responses carry the data only and the trace is in the JSONL audit log on the server side.
/rulake-query "what does ADR-157 commit to?"
/rulake-query --intent verify --freshness_ms 0 "is bundle b1f4...3d valid?"
/rulake-query --target_collection adrs.rulake "list all merged ADRs"
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Use ruLake as the agent's working memory. Auto-invoke /rulake-memory:memory-recall when the agent needs to look up something it might already know; auto-invoke /rulake-memory:memory-remember after the agent learns or decides something worth pinning. Closes the self-learning loop without requiring the agent author to wire it explicitly.
中文简介正在准备中。当前显示原文说明。