Unforget

Fast, self-hosted memory for AI agents. Zero-LLM writes. PostgreSQL + pgvector.

pip install unforget

~7ms writes~25ms recall0 LLM calls on write4 retrieval channels

Quick Start


from unforget import MemoryStore
 
store = MemoryStore("postgresql://user:pass@localhost/db")
await store.initialize()
 
memory = store.bind(org_id="acme", agent_id="support-bot")
 
# Write — instant, no LLM
await memory.write("User prefers dark mode", tags=["preference"])
 
# Recall — 4-channel hybrid search
results = await memory.recall("user preferences", limit=5)
for r in results:
    print(f"{r.content} (score: {r.score:.3f})")

Why Unforget?


~7ms writes	Zero LLM calls on the write path. Embed + insert, done.
~25ms recall	Semantic + BM25 + entity + temporal search, fused with RRF.
Self-hosted	PostgreSQL + pgvector. No cloud vendor, no vector DB dependency.
Background consolidation	Dedup, decay, and LLM promotion while agents sleep.
Drop-in wrappers	`wrap_openai()` and `wrap_anthropic()` — memory in one line.
Apache 2.0	Open source, no vendor lock-in.

Get Started → · GitHub →