# OpenOpenAI

> OpenOpenAI is a self-hosted, open-source OpenAI Assistants API implementation wire-compatible with official SDKs for custom models and RAG.

- **Category:** AI Assistant
- **Licence:** MIT
- **Open source:** oss
- **Languages:** Typescript
- **Platforms:** Node.js
- **Tags:** Redis, Ai Assistant, LLM, Rag, Openai, Typescript, S3, Self Hosted, Openai Api, Assistants api, Postgres, Assistants, Gpts

## Description

OpenOpenAI implements the stateful portions of OpenAI's Assistants API — Assistants, Threads, Runs, Messages, Files, and the retrieval tool — with 100% auto-generated API routes and types from OpenAI's OpenAPI spec. This means you can point the official OpenAI Node.js or Python SDK at your own server by only changing the baseURL. The project consists of a REST API server (Hono) and an async task runner (BullMQ) that process thread runs. It requires Postgres as the primary datastore via Prisma, Redis for the task queue, and S3-compatible storage (like Cloudflare R2, MinIO, or LocalStack) for file uploads. An OpenAI API key is currently required for the underlying chat completion calls, though support for local models is planned. The server listens on port 3000 by default. The project is MIT licensed and actively developed.

## Links

- [Homepage](https://github.com/transitive-bullshit/OpenOpenAI)
- [GitHub Repository](https://github.com/transitive-bullshit/OpenOpenAI)
- [Author Website](https://transitivebullsh.it)

---
Source: https://selfhost.directory/project/openopenai