Expand description
§MemoryHub
A centralized memory service for AI agents.
Personal AI agents (Claude Code, etc.) each build up knowledge as local Markdown files, in isolation. MemoryHub pools those per-user memories into one shared, searchable knowledge base: it ingests raw Markdown from each agent, embeds it for semantic search, and runs a synthesizer that produces per-user and cross-user summaries.
§What it does
- Collects memories from many users and agents over a small JSON API.
- Searchable instantly — every memory is indexed the moment it’s written, so it’s immediately available. No batch or reprocess step.
- Finds the right memory — search handles both meaning-based and exact keyword queries, so results hold up whether you describe an idea or quote a term.
- Synthesizes — generates per-user and cross-user summaries from the pooled memories as they come in.
- Human-readable storage — memories stay as plain Markdown files you can read, edit, and version directly.
- Bring your own models — chat and embeddings are configured separately, so you can mix providers (and swap in a stub for tests).
It’s a single self-contained service.
Re-exports§
pub use config::Config;
Modules§
- config
- error
- Error types for MemoryHub.
- http
- HTTP service: Axum-based JSON API over the Memory Manager actor.
- llm
- LLM Service Actor — the single entry point for embedding and synthesis.
- memory
- Memory actor sub-system for MemoryHub.