engram-rs

Engram-rs is a Rust-based memory engine for AI agents. It stores memories in three layers (Buffer, Working, Core) and uses an automatic decay mechanism, a quality gate for long-term retention, and a self-organizing topic tree to keep knowledge approachable. The server exposes a lightweight HTTP API that lets you store memories, recall by meaning, resume sessions, and interact with triggers and topics. Typical workflows include storing a memory with content and tags, recalling relevant memories by a query, and resuming a session to restore full context. You can also browse and expand topics, and use triggers to fetch groups of relevant lessons before risky operations.

Prerequisites:

• Docker installed and running
• Basic familiarity with curl or any HTTP client

Installation steps (Docker):

1. Pull and run the Engram image:

docker pull ghcr.io/kael-bit/engram-rs:latest
docker run -d --name engRAM -p 3917:3917 ghcr.io/kael-bit/engram-rs:latest

2. Verify the service is running. You should be able to reach the API at http://localhost:3917/

3. Interact with the API (examples below) to store, recall, or resume:

curl -X POST http://localhost:3917/memories \
  -d '{"content": "Always run tests before deploying", "tags": ["deploy"]}'

curl -X POST http://localhost:3917/recall \
  -d '{"query": "deployment checklist"}'

4. When you no longer want the container, stop and remove it:

docker stop engRAM
docker rm engRAM

Alternative: Build from source (Rust) if you prefer to run locally without Docker:

• Ensure Rust toolchain is installed (Rust 1.75+)
• Clone the repository and build the binary
• Run the binary directly with the appropriate flags (see repository docs for exact runtime arguments).

Tips:

• The server assumes port 3917 by default (as shown in examples). If you change the port, update your requests accordingly.
• The API includes endpoints for memories, recall, resume, and topic/navigation features (e.g., POST /memories, POST /recall, GET /resume, POST /topic).
• The memory system emphasizes a balance between recall performance and retention; tune your usage by adjusting tags and the content you store.
• When deploying with Docker, consider mapping persistent storage if you want memory data to survive container restarts (mount a host directory to the container path used by Engram).
• If you encounter network or port issues, ensure no other service is occupying 3917 and that Docker has access to the host network as needed.