An OpenAI-compatible API that currently only speaks GPT4All

What it does

GenossGPT is a Python server that mimics the OpenAI ChatGPT and Embeddings API, translating requests to run on local open-source models instead of OpenAI’s servers. At the moment, the only supported backend is GPT4ALL, so the “all LLMs” promise in the description is largely aspirational. It was originally extracted from the Quivr project when its model-abstraction logic grew unwieldy.

The interesting bit

The clever part is the API compatibility: because Genoss speaks the same JSON shape as OpenAI’s endpoints, existing clients can talk to it without rewrites. That makes it glue code, but useful glue if you are trying to de-cloud an existing chatbot.

Key highlights

• Exposes an OpenAI-compatible REST API for chat completions and embeddings
• Currently routes only to GPT4ALL; HuggingFace and cloud providers are on the roadmap
• Runs fully offline via local model binaries
• Bootstrapped from the Quivr project to isolate LLM abstraction complexity
• Apache 2.0 licensed

Caveats

• Only GPT4ALL is actually supported today, despite the ambitious “all LLMs” branding
• The project is in a pre-publish state and requires manually building GPT4ALL backend binaries
• Several roadmap items (user management, cloud providers, multi-modal models) are unchecked

Verdict

Worth a look if you need a local, offline drop-in for OpenAI’s chat API and are willing to tolerate early-stage rough edges. Skip it if you need broad model support today or a managed service.