yantrikdb-server

Multi-tenant cognitive memory database server. Wire protocol + HTTP gateway, built-in embeddings, automatic replication and failover.

Built on top of yantrikdb — the cognitive memory engine.

What it is

YantrikDB is the memory subsystem for AI agents. Not a vector store, not a key-value database — a cognitive memory database that:

• Decays — memories fade unless reinforced (exponential half-life)
• Consolidates — similar memories auto-merge
• Conflicts — contradictions are detected and surfaced
• Sessions — conversation context is first-class
• Explains — recall tells you why each result was returned
• Personalizes — emergent personality traits from access patterns
• Replicates — multi-node CRDT-based with auto failover
• Encrypts — AES-256-GCM at rest, TLS in transit

Install

cargo install yantrikdb-server

Or download a pre-built binary:

wget https://github.com/yantrikos/yantrikdb-server/releases/latest/download/yantrikdb-linux-amd64
chmod +x yantrikdb-linux-amd64

Quick start

yantrikdb db --data-dir ./data create default
yantrikdb token --data-dir ./data create --db default
yantrikdb serve --data-dir ./data
# Wire protocol on :7437, HTTP on :7438

Cluster setup

yantrikdb cluster init \
  --node-id 1 \
  --output ./node1.toml \
  --peers 192.168.1.2:7440 \
  --witnesses 192.168.1.3:7440
yantrikdb serve --config ./node1.toml

See the main README for full deployment guides, HTTP API reference, and operational documentation.

Companion crates

• yql — interactive REPL client (like psql)
• yantrikdb-witness — vote-only daemon for 2-node failover
• yantrikdb-protocol — wire protocol codec

License

AGPL-3.0-only — server and core engine. Client SDKs (yql, Python SDK) are MIT-licensed.