Annal

关于

Annal functions as a persistent semantic memory server, offering AI agent teams a shared institutional memory that survives across sessions. It helps prevent contradictions and preserves crucial context that might otherwise be lost, ensuring decisions made months ago surface automatically when relevant. The server utilizes pluggable vector backends like ChromaDB or Qdrant, employs local ONNX embeddings for efficiency, and is exposed as an MCP server for seamless integration into multi-agent workflows.

主要功能

• Optional file indexing for directories, chunking markdown by heading, and incremental re-indexing via watchdog events.
• Hybrid search (with Qdrant) combining dense vector similarity with BM25 keyword matching via reciprocal rank fusion.
• 0 GitHub stars
• Web dashboard for memory browsing, semantic search, bulk deletion, and live updates.
• Comprehensive tools for storing, searching, updating, and managing memories, including near-duplicate detection and supersession tracking.
• Persistent MCP server with pluggable vector backends (ChromaDB, Qdrant) and local ONNX embeddings.

使用案例

• Equipping AI agents with long-term memory to load context at session start or before diving into unfamiliar code.
• Storing and retrieving bug root causes, architectural decisions, and codebase patterns for future reference.
• Facilitating cross-session continuity for multi-agent teams by providing shared institutional knowledge and preventing repeated investigations.