cmdcode2apipy

cmdcode2apipy is a small OpenAI-compatible gateway for commandcode CLI. It lets OpenAI-style clients call commandcode.ai models through familiar endpoints such as /v1/chat/completions and /v1/models.

Made specifically for commandcode Go 1 USD/m plan

An adapter like this is required as the cheapest 1 USD/m commandcode.ai Go plan does not provide direct access to it's OpenAI-compatible API: you have to host an adapter like this on your own.

Features

• OpenAI-compatible HTTP API
  • POST /v1/chat/completions
  • GET /v1/models
• Streaming and non-streaming chat completions
• OpenAI image_url input conversion to commandcode.ai / Anthropic-style image blocks
• Local bearer-token auth for clients
• CORS enabled for local UI clients
• Usage counter persisted to usage.json
• Health endpoint: GET /health
• Usage endpoint: GET /usage

Setup & Run

1. Clone the project:

   git clone https://git.blackruby.hu/blackruby/cmdcode2apipy.git && cd cmdcode2apipy
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Configure the gateway: Create a config.yaml. The app will auto-generate a default one on first run if missing.

   api_key: "ccgw-your-local-client-key"
   commandcode:
     api_key: "your-command-code-api-key"
     base_url: "https://api.commandcode.ai"
   host: "localhost"
   port: 11434
   exclude_models:
     - "gpt-"
     - "claude-"
     - "gemini-"
   

   Note: You can obtain your commandcode.ai API key through the commandcode.ai CLI or dashboard.

4. Run the server:

   python app.py
   

   (Alternatively, use uvicorn directly: uvicorn app:app --host localhost --port 11434)

5. Run tests:

   pytest tests/ -v
   

Project Layout

python/
├── app.py                 # FastAPI application entry point
├── config.yaml            # Gateway configuration
├── requirements.txt       # Python dependencies
├── src/
│   ├── __init__.py
│   ├── config.py          # Config loading and generation
│   ├── models.py          # Pydantic models (OpenAI & CC formats)
│   ├── client.py          # Async httpx client for CC API
│   ├── translation.py     # OpenAI <-> CC request/response mapping
│   ├── handlers.py        # FastAPI routes, middleware, and SSE logic
│   └── usage.py           # Thread-safe usage tracking
└── tests/                 # pytest test suite

Configuration

config.yaml is created automatically and intentionally ignored by git.

Fields:

• api_key — local bearer token required by clients calling this gateway.
• commandcode.api_key — Your commandcode.ai API key.
• commandcode.base_url — commandcode.ai API base URL.
• host — HTTP listen host. Defaults to localhost. Use 0.0.0.0 to listen on all interfaces.
• port — local listen port. Defaults to 11434.
• exclude_models — model ID prefixes hidden from /v1/models and rejected by /v1/chat/completions.

New configs exclude gpt-, claude-, and gemini- by default as commandcode.ai Go plan does not provide access to them. These prefixes match both plain model IDs such as gpt-4 and provider-qualified IDs such as openai/gpt-4 by checking the part after the final /.

To make all models available, remove the entries or set an empty list:

exclude_models: []

Use with OpenAI-compatible clients

Set the base URL to your local gateway:

http://localhost:11434/v1

Use the generated api_key from config.yaml as the bearer token.

curl example

curl http://localhost:11434/v1/chat/completions \
  -H "Authorization: Bearer <local-api-key>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek/deepseek-v4-pro",
    "messages": [
      {"role": "user", "content": "Hello!"}
    ],
    "stream": false
  }'

Streaming

curl http://localhost:11434/v1/chat/completions \
  -H "Authorization: Bearer <local-api-key>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek/deepseek-v4-pro",
    "messages": [
      {"role": "user", "content": "Write a short haiku."}
    ],
    "stream": true
  }'

Endpoints

GET /health

No authentication required.

{"status":"ok"}

GET /usage

No authentication required. Returns locally accumulated usage counters:

{
  "total_requests": 1,
  "prompt_tokens": 7527,
  "completion_tokens": 55,
  "cache_read_tokens": 7424,
  "cache_write_tokens": 0
}

Usage is persisted to usage.json, which is ignored by git.

GET /v1/models

Returns the model list after applying exclude_models filtering.

POST /v1/chat/completions

Accepts OpenAI-style chat completion requests and forwards them to commandcode.ai. Requests for excluded models return 404 with the existing OpenAI-compatible error JSON shape.

Supported request styles:

• Plain text messages
• Multimodal content arrays with image_url
• stream: true server-sent events
• stream: false JSON response

Notes

The project is a Python rewrite of cmdcode2api originally written in Go, which is no longer actively maintained nor is accepting contributions.

This is gateway currently targets the commandcode.ai API shape observed during development. If commandcode.ai changes its internal API, the adapter may need updates.