fathomdb 0.2.5

Local datastore for persistent AI agents — graph, vector, and full-text search on SQLite

fathomdb

Local datastore for persistent AI agents — graph, vector, and full-text search on SQLite.

Installation

pip install fathomdb

Quick Start

from fathomdb import Engine

engine = Engine.open("my_agent.db")

with engine.session() as session:
    # Write data
    w = session.write_builder()
    node = w.insert_node(kind="memory", properties={"text": "hello world"})
    session.execute_write(w)

    # Query data
    q = session.query_builder()
    q.kind("memory")
    results = session.execute_query(q)

engine.close()

Features

• Graph backbone with nodes, edges, and temporal tracking
• Full-text search via SQLite FTS5
• Vector similarity search via sqlite-vec
• Provenance tracking on every write
• Single-writer / multi-reader with WAL

Documentation

See the GitHub repository for full documentation.

License

MIT