nautilus-compass

chunxiaoxx/nautilus-compass is ranked 10,949 among these 8 peers; ChristopherKahler/paul is ranked 10,947.

nautilus-compass **Open-source memory & reliability layer for AI agents. Plugin for Claude Code / Desktop · Cline · Cursor · Continue. Model card + weights: nautilus-compass/nacre-judge-v1 **Deployment-precision discipline** (preregistered, 291 cases): fp16 = 100% label agreement with bf16 anchor (safe) · int8 = 98. 6pp; dynamic-weight routing suspended until corpus ≥3,000. Negative results ship as first-class findings ( ). ** Raw text embedded locally with BGE-m3. No extraction LLM at ingest, no graph, no data leaving your machine. In Aug 2026

we added **utterance-routed chunk retrieval**: single-session and knowledge-update questions route to turn-window chunks (the answer usually lives in ONE user turn; whole-session embedding dilutes it), everything else uses session-level hybrid (BM25 + dense RRF). Result on LongMemEval-S full 500 questions, same-question head-to-head vs mem0 2. 834 | One-command reproduction (retrieval-only, no LLM calls, CPU works / GPU recommended for the full 500): The same utterance ammo overtakes mem0 **on its own

home benchmark** (LOCOMO-10, n=1986: 0. 802) and fixes the single-session collapse on LongMemEval-M (0. ** Memory recalled doesn't stop the AI from breaking the rule *this time*. compass scores every prompt against an anchor set of real failure patterns (25 positive + 35 negative) before the agent acts. Only set it if you know why; the daemon logs a WARNING when it's off. - Token changes to hot-reload via mtime check (no systemd restart needed since v3. - First

recall after daemon idle may take up to 90 s (model cold-load); the MCP client auto-retries once with the extended timeout. The installer wires three hooks into : → time-bucketed memory recall + drift check → mid-session writer → end-of-session summary (writes a session battle-report to ) Slash commands: · · · · . Any other MCP client Auto-detects Claude Desktop, Cursor, Cline, Continue. Per-agent copy-paste configs: · raw protocol: . Scopes are server-bound to your own space (read+write); cross-user

access is denied and verified by automated probes. MCP endpoint: (Bearer token · streamable-http). Add to install the fused status line (live recall hit-counter 🧠, drift state, 5-min traffic). 834, our reproduction, each side on its own default embedder): +11. Largest flip: single-session-user P@1 0. 49 | | **LOCOMO-10** (n=1986 · mem0's home benchmark) | **P@1 0. 8pt | | **LongMemEval-M 500q full** (~501 sessions/question) | **P@5 0. 888** | 12x larger session pools cost only 9pt vs

S500; ssu collapse fixed at n=500 (0. 53 newly exposed; no mem0 M head-to-head yet | | **EverMemBench-Dynamic** (n=500) | **44. 3% (Run 2)** | tops the four published Table 4 baselines (Mem0 37. 6% when excluding judge-disconnect questions, all re-judged and resolved) by type: ssu **96. Cross-session types jumped +45~60pt (ms 22. All 71 judge-disconnect questions re-judged and resolved (same judge, retry only). Full verdict: ; baseline JSON: | | **LongMemEval-V2** (official benchmark by

xiaowu0162, 451q agent-trajectory memory · our tuned run, rejudged clean 2026-09-02) | **web 40. 1% without trajectory evidence) | brand