LM Studio setup

Some extensions can use a second model for reviewer or sub-agent tasks. LM Studio is the recommended local model server for this.

Basic setup

  1. Download and install LM Studio
  2. Load a model (e.g. Qwen2.5-Coder-7B-Instruct)
  3. Start the local server: Local Server tab → Start Server (default: http://localhost:1234)

Configuring pi to use LM Studio

In your pi settings (~/.pi/agent/settings.json or .pi/settings.json):

{
  "providers": {
    "lmstudio": {
      "baseUrl": "http://localhost:1234/v1"
    }
  }
}

dual-review with a second model

Set the DUAL_REVIEW_MODEL environment variable to the model name loaded in LM Studio:

# In ~/.pi/agent/.env
DUAL_REVIEW_MODEL=qwen2.5-coder-7b-instruct

The dual-review extension will route /review and dual_review tool calls to that model.

Use case Model Size
Reviewer (dual-review) Qwen2.5-Coder-7B-Instruct ~5 GB
Reviewer (faster) Qwen2.5-Coder-3B-Instruct ~2 GB
Subagent Any code-capable model —

Notes

  • LM Studio must be running before starting pi with these extensions enabled
  • The primary model (used for the main agent) is configured separately in pi's provider settings
  • Verify connectivity: curl http://localhost:1234/v1/models