Pi

Pi is a minimal, highly extensible terminal coding framework with custom skills, templates and themes. It connects to any OpenAI-compatible provider via models.json.

• Repository: https://github.com/badlogic/lemern/tree/main/packages/pi-mono

Prerequisites

• Pi installed
• Base URL: https://api.tokenhub.belivestudios.com
• API Key: create one in "API Tokens" on the BeLive Studio Console
• Model name: exactly matching a model ID in the console model list

Edit models.json

Edit ~/.pi/agent/models.json and add the BeLive Studio provider:

{
  "providers": {
    "belive-studios": {
      "baseUrl": "https://api.tokenhub.belivestudios.com",
      "api": "openai-completions",
      "apiKey": "$BELIVE_STUDIOS_API_KEY",
      "models": [
        { "id": "gpt-5.6-sol", "name": "GPT-5.6 Sol (BeLive Studio)" },
        { "id": "claude-fable-5", "name": "Claude Fable 5 (BeLive Studio)" },
        { "id": "gemini-3.7-flash", "name": "Gemini 3.7 Flash (BeLive Studio)" }
      ]
    }
  }
}

apiKey supports $ENV_VAR interpolation, so export the key first:

export BELIVE_STUDIOS_API_KEY="sk-your-belive-studios-key"

Start Using

cd /path/to/your/project
pi

Type /model and pick a model under the belive-studios provider.

FAQ

BeLive Studio models not shown in /model? Pi only lists models whose provider has credentials — make sure $BELIVE_STUDIOS_API_KEY is set, or save a key with /login.

Errors with reasoning models? Some compatible servers reject the developer role or reasoning_effort parameter. Add this to the provider config:

"compat": { "supportsDeveloperRole": false, "supportsReasoningEffort": false }