jsonbored/mem0-aio

mem0-aio is a packaging project that bundles an open source AI memory layer called Mem0 (OpenMemory) into a single Docker container, with a template aimed at Unraid home server users. Mem0 is a service that lets AI agents and language models remember things across conversations, by storing notes in a vector database and retrieving them when relevant. This repo's job is to make the easiest possible self-hosted install on day one: the container already includes the Qdrant vector database and persistent local storage, plus the OpenMemory web dashboard on port 3000 and its API/MCP server on port 8765 inside the container. The README walks through a beginner install path for Unraid. You install the supplied Community Applications template, keep the default appdata folder, set either an OpenAI API key for the hosted quick start or an OLLAMA_BASE_URL pointing at your local Ollama server for a fully local setup, choose which LLM and embedding model to use, start the container, and open port 3000 in your browser. The wrapper auto-detects the embedding size needed by Qdrant for normal Ollama models. The API/MCP port is intentionally bound to localhost inside the container by default, so the memory API and the model credentials it holds are not accidentally exposed to the network. Direct clients that want to connect to port 8765 must explicitly publish the port and switch the bind address to 0.0.0.0, which the README presents as a deliberate security decision rather than a missing feature. For power users, the advanced template exposes the full upstream OpenMemory environment surface: alternative LLM providers like Anthropic, Groq, Together, DeepSeek, Azure OpenAI, and Bedrock-compatible providers, and alternative vector stores including Chroma, Weaviate, Redis, pgvector, Milvus, Elasticsearch, OpenSearch, and FAISS. Only one external vector backend can be active at a time. The repo also documents the release flow, publishes to Docker Hub, and has 15 stars.