trvon/yams

Persistent memory for LLMs and apps. Content-addressed storage with dedupe, compression, full-text and vector search.

What it solves

YAMS provides a persistent, searchable memory system for code, documents, and local applications. It addresses the problem of efficiently storing and retrieving information from a large local corpus of data without requiring a hosted service or account.

How it works

The system uses SHA-256 content-addressed storage with chunk-level deduplication and compression to store data. For retrieval, it combines multiple search surfaces: text (SQLite FTS5), vector search (Simeon embeddings), hybrid search, and knowledge-graph retrieval. For code specifically, it uses Tree-sitter for symbol extraction and source-aware graph traversal.

Who it’s for

Developers and users who need a local-first, persistent memory system for their files and codebases, as well as those who using MCP-compatible developer tools.

Highlights

  • own local-first operation with no account requirements
  • Content-addressed storage with deduplication and compression
  • Hybrid search capabilities including text, vector, and graph retrieval
  • MCP server support for integration with developer tools
  • Tree-sitter symbol extraction for source-aware code analysis
  • Built-in corruption detection and repair tooling

Related

  • Project
    ThinkfleetAI/memmesh

    A persistent, local-first memory store for AI agents that provides durable, typed, and searchable memory using a single Rust binary without requiring a hosted vector database.

  • Project
    GeminiLight/MindOS

    A local-first knowledge base that provides a shared memory and durable context for humans and multiple AI agents via MCP and Skills.

  • Project
    andrea9293/mcp-documentation-server

    A local-first document management and semantic search server for AI agents that provides a private knowledge base with a built-in web UI and hybrid search.

  • Project
    facebookresearch/faiss

    A library for efficient similarity search and clustering of dense vectors, capable of handling billion-scale datasets using CPU and GPU acceleration.

  • Project
    tickernelz/opencode-mem

    A persistent memory system for AI coding agents that uses a local vector database to retain long-term technical context and user preferences across sessions.