oc-memory
♥Cherished
A Rust-based local memory engine that integrates with OpenClaw via MCP protocol, enabling session-persistent retrieval and storage of information without external API dependencies. It combines BGE-m3-ko embeddings with BM25 search optimized for Korean text, reducing token usage through progressive disclosure patterns.
Clauded With Love Rating
8.2 / 10
A Rust-based local AI memory engine that integrates with OpenClaw via MCP protocol, combining BGE-m3-ko embeddings with BM25 search for Korean text optimization. The system enables session-persistent memory storage and retrieval without external API dependencies, using progressive disclosure patterns to reduce token usage.
Code Quality8.2
Usefulness7.8
Claude Usage8.5
Documentation9.1
Originality7.4
Highlights
- ✓Comprehensive architecture with 6 well-structured crates and proper separation of concerns (core, embeddings, search, observer, mcp-server, server)
- ✓Excellent documentation with detailed installation scripts, one-liner setup for cloud environments, and clear integration examples for OpenClaw
- ✓Sophisticated hybrid search scoring system combining semantic similarity (0.6), keyword matching (0.15), recency decay (0.15), and importance weighting (0.10)
To Improve
- →Add error recovery mechanisms and graceful degradation when embedding models fail to load or ONNX runtime encounters issues
- →Implement memory compaction and archival strategies to prevent unbounded growth of the local storage over time