DevTrendsDeveloper trends
Accent color
Weekly Submissions ↗

chunxiaoxx /

nautilus-compass

Open-source memory layer for AI agents: zero-token writes via local BGE-m3, question-routed turn-level retrieval. LongMemEval-S P@1 0.890 vs mem0 0.774. Source-tagged governance, shared memory capsules, signed reproducib…

DTSoftware developmentDTAgent extensionDTMCP serviceDTLibrary / SDKDTClaude CodeDTCursorLanguagePythonOriginala2aOriginalagent-memoryOriginalagentic-aiOriginalclaude-codeOriginalcross-agent
GitHub repository Website

Product screenshots

About the project

Open-source memory layer for AI agents: zero-token writes via local BGE-m3, question-routed turn-level retrieval. LongMemEval-S P@1 0.890 vs mem0 0.774. Source-tagged governance, shared memory capsules, signed reproducible evidence.

Descriptions are drawn from collected community sources.

Discovery history

Dates refer to DevTrends reports. Metrics reflect the snapshot collected at the time.

  1. ↗

    Open-source memory layer for AI agents: zero-token writes via local BGE-m3, question-routed turn-level retrieval. LongMemEval-S P@1 0.890 vs mem0 0.774. Source-tagged governance, shared memory capsules, signed reproducible evidence.

    Weekly Submissions ↗

DEV TRENDS / COMMUNITY

Discuss this project

Sign in with GitHub to share experience, questions, or useful project updates.

Open discussions ↗