Use a self-hosted model
Run Pullwise reviews on your own hardware with Ollama, LM Studio, vLLM or llama.cpp so your diffs never reach a third-party AI provider.
Pullwise can send reviews to any server that speaks the OpenAI chat-completions API. Inference then runs on your hardware and the diff never reaches a third-party AI provider.
Supported runtimes
Anything that serves /v1/chat/completions, including Ollama, LM Studio, vLLM and llama.cpp’s server.
Recommended model
We recommend qwen2.5-coder:7b as a starting point. It runs on a single modern GPU or an Apple silicon Mac. Use qwen2.5-coder:14b or larger for more thorough reviews.
Example: Ollama
ollama pull qwen2.5-coder:7b
ollama serveThen in AI, select Add key on the Local / Ollama card and set the fields below. Save, then select Verify and Make active.
| Setting | Example |
|---|---|
| Base URL | https://ollama.internal.example.com/v1 |
| Model | qwen2.5-coder:7b |
| API key | Optional, if your server requires one |
Network access
Pullwise’s servers need to reach your endpoint over HTTPS. Put it behind a reverse proxy with TLS and an API key, and allow only Pullwise’s traffic. If your code must never leave your network at all, ask us about a private deployment.