pi-sync

Keep Pi coding agent's models.json in sync with your local vLLM or llama.cpp llama-server.

pi-sync queries GET /v1/models, updates ~/.pi/agent/models.json, and preserves existing model metadata for IDs that are still served. If models.json does not exist, it creates a local llama-cpp provider using the OpenAI-compatible completions API.

Install

pip install pi-codingagent-sync

For local development:

pip install -e .

Usage

Sync against the URL already in models.json, or create the default local config:

pi-sync

Point at a specific server and update the stored URL:

pi-sync --host 127.0.0.1 --port 8080

pi-sync also respects llama.cpp's server target environment variables:

LLAMA_ARG_HOST=127.0.0.1 LLAMA_ARG_PORT=8080 pi-sync

Target precedence is:

CLI --host/--port values override LLAMA_ARG_HOST/LLAMA_ARG_PORT.

Preview changes without writing:

pi-sync --dry-run

Auto-sync on every Pi launch

Install the wrapper once:

pi-sync install

This writes ~/.local/bin/pi, which runs pi-sync silently and then execs the real pi binary. Make sure ~/.local/bin appears earlier in PATH than the real Pi binary.

Default Config

When creating a new config, pi-sync starts with:

{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "llama",
      "compat": {
        "supportsDeveloperRole": false,
        "supportsReasoningEffort": false
      },
      "models": [
        {
          "id": "qwen3-coder-local",
          "name": "Qwen local via llama-server",
          "input": ["text"],
          "contextWindow": 32768,
          "maxTokens": 8192
        }
      ]
    }
  }
}

Options

pi-sync [--host HOST] [--port PORT] [--provider ID] [--config PATH]
        [--dry-run] [--no-url-update] [--timeout SECONDS]

The default config path is $PI_CODING_AGENT_DIR/models.json, or ~/.pi/agent/models.json when PI_CODING_AGENT_DIR is unset.

Development

pip install -e ".[dev]"
python -m pytest