EverMemOS

EverMemOS is a memory-powered AI infrastructure that provides long-term memory for conversational agents. It exposes a REST API-backed platform and integrates with vector stores and databases (Milvus, Elasticsearch, MongoDB) to store episodes, memories, and relationships. The server is designed to be run via the uv package manager, enabling developers to deploy EverMemOS quickly and connect agents or plugins that require persistent memory, retrieval-enhanced reasoning, and memory-augmented interactions. Typical workflows include creating and querying memory episodes, retrieving related memories through BM25 or embeddings, and leveraging agent memory to inform decision-making across sessions. The system is built to work with a variety of tools and plugins, enabling platform integrations (e.g., VSCode, Chrome, Notion) and agent architectures that rely on long-term context.

To interact with the server, you’ll use its REST API endpoints for storing and retrieving memories, managing profiles, and performing advanced retrieval. The server integrates with Milvus for vector embeddings, Elasticsearch for searchable indexing, and MongoDB for document storage. You can run your own agent that reads conversations, encodes memories, and queries EverMemOS for relevant context to improve responses, planning, and decision-making. If you’re building an agent with long-term memory, EverMemOS provides memory episodes, facts, and relations that your agent can reference across sessions.

Prerequisites

• Python 3.10+
• Docker 20.10+ (optional for containerized deployment)
• uv package manager (uvx) for Python deployments

Install and run locally with uvx

1. Ensure Python is installed and accessible:

python3 --version

2. Install the uv package manager if you don’t have it yet:

pip install uvx

3. Install the EverMemOS package (replace with the exact package name if different):

uvx install evermemos

4. Run EverMemOS using uvx (start the server):

uvx run evermemos

Optional: Run with Docker

1. Pull the image (if an official image is published):

docker pull evermemos/evermemos:latest

2. Run the container (example):

docker run -p 8000:8000 -e MONGO_URI=mongodb://localhost:27017/evermemos evermemos/evermemos:latest

Prerequisites recap: ensure Python 3.10+, Docker installed if using containers, and uvx for Python deployments.

Tips and common issues:

• Environment variables: set MONGO_URI, ELASTICSEARCH_URL, and MILVUS_HOST as needed to point to your databases.
• If your memory database (Milvus) is not reachable, verify network settings and credentials.
• For large-scale deployments, consider container orchestration and dedicated storage for memory indexes.
• Check API docs for available endpoints: memory episodes, memories, profiles, and retrieval options (BM25 vs embeddings).
• When upgrading, review release notes for API changes or migration steps, especially around schema changes in memory objects.
• Ensure Python 3.10+ compatibility and that uvx is the active package manager in your environment.