Mem0 API

This project provides a production-ready API for mem0, a tool for AI-powered semantic memory management. It allows you to store, search, and manage memories for AI agents using a simple, robust interface.

This API is designed for developers who need to give their AI agents a reliable long-term memory. It handles the complexities of vector embeddings and LLM integration, so you can focus on building your agent's logic.

The Problem It Solves

AI agents are often stateless. They forget everything after an interaction. This makes building sophisticated, context-aware agents difficult.

mem0 solves this by providing a memory layer. This API exposes that layer through a clean, secure, and scalable service. It's built for real-world use, not just as a demo.

Getting Started

You need Python 3.9+ and access to a Redis instance. You will also need an API key from an LLM provider like Google Gemini or OpenAI.

1. Clone the repository and install dependencies:

   git clone https://github.com/OctavianTocan/mem0-api.git
   cd mem0-api
   pip install -r requirements.txt

2. Configure your environment. Copy the example .env.example file to .env and fill in the required values. At a minimum, you need to set MEMORY_API_KEY.

   cp .env.example .env

3. Run the server.

   ./start.sh

   The API will be available at http://localhost:8000.

How to Use It

All endpoints require an X-API-Key header for authentication. This key should match the MEMORY_API_KEY you set in your .env file.

Adding a Memory

To store information, send a POST request to /add_memory. The messages field contains the content to be stored.

POST /add_memory
Content-Type: application/json
X-API-Key: your-secret-key

{
  "messages": [
    {"role": "user", "content": "The user's favorite food is pasta."}
  ],
  "user_id": "user-123"
}

The infer parameter, which is on by default, uses the LLM to extract and store the core facts from the content.

Searching Memories

To retrieve information, send a POST request to /search_memory. The API performs a semantic search, so you can ask questions in natural language.

POST /search_memory
Content-Type: application/json
X-API-Key: your-secret-key

{
  "query": "What is the user's favorite food?",
  "user_id": "user-123"
}

The API will return the most relevant memories it has stored for that user.

Limitations

This API is a starting point. It provides core memory functions but is not a complete agent framework. It manages memory, not agent state or decision-making.

The quality of the semantic search depends heavily on the underlying embedding model. The default models are a good starting point, but you may need to experiment to find the best fit for your use case.

Project Roadmap

This project is under active development. Future plans include:

• Support for more vector database backends.
• More sophisticated memory analysis and summarization features.
• Integration with popular agent frameworks.

Contributions are welcome.

Project Information

This project is licensed under the MIT License. See the LICENSE file for details.

All contributors are expected to follow our CODE_OF_CONDUCT.md. Please read it before participating. Our docs/CODING_GUIDELINES.md outlines the standards for code contributions.