Skip to content
ThinkReviewThinkReview Docs

Local Ollama

ThinkReview can run AI code reviews against Ollama on your machine. Patches are sent to your local models instead of ThinkReview cloud—useful for air-gapped environments and privacy-sensitive codebases.

Prerequisites

Section titled “Prerequisites”
  1. Ollama installed and running locally.
  2. At least one model pulled (for example llama3, qwen2.5-coder, or another model your hardware can run).
  3. The ThinkReview browser extension installed.

Allow the extension to reach Ollama

Section titled “Allow the extension to reach Ollama”

Browsers block cross-origin requests to localhost unless Ollama allows them.

  1. Stop any running Ollama process.
  2. Restart Ollama with CORS open to the extension, for example:
Terminal window
OLLAMA_ORIGINS="*" ollama serve

On some setups you may set OLLAMA_ORIGINS permanently in your shell profile or service environment.

Configure ThinkReview

Section titled “Configure ThinkReview”
  1. Open the ThinkReview extension settings.
  2. Choose Ollama as the provider (alongside Cloud AI / other options).
  3. Select a model that is already pulled in Ollama.
  4. Save, then open a pull or merge request and run a review.

Tips

Section titled “Tips”
  • Keep Ollama running while you review; if the daemon stops, reviews fail with a connection error.
  • For team-wide self-hosting with a shared LLM endpoint (not only laptop Ollama), see Self-Hosted Enterprise Gateway.
  • Cloud models remain available when you switch the provider back—see Model Selection and available models.
Section titled “Related”

TL;DR: Point ThinkReview at local Ollama (OLLAMA_ORIGINS + provider setting) so PR reviews run offline on your machine.