Engram Documentation

Engram is a decentralized vector database on Bittensor — permanent, content-addressed semantic memory for AI applications. No central server, no single point of failure.

What is Engram?

Engram applies the IPFS insight to AI memory: every piece of knowledge gets a CID derived deterministically from its embedding. The same text always maps to the same CID — regardless of which miner stores it.

from engram.sdk import EngramClient
client = EngramClient("http://127.0.0.1:8091")
# Store text — returns a permanent CID
cid = client.ingest("The transformer architecture changed everything.")
print(cid)  # v1::a3f2b1c4d5e6f7a8b9c0d1e2f3a4b5c6
# Semantic search — no exact match needed
results = client.query("how does attention work?", top_k=5)
for r in results:
    print(f"{r['score']:.4f}  {r['cid']}")

How it works

Network