Turn DuckDuckGo's free chat into your private OpenAI API

What it does

Duck2api is a lightweight Go server that proxies chat requests to DuckDuckGo’s AI service and returns results through a standard OpenAI-style /v1/chat/completions endpoint. It includes a basic web interface and runs as a standalone binary or container. The goal is to let you query DuckDuckGo’s hosted models using the same client code you’d use for OpenAI.

The interesting bit

The project is essentially a protocol translator that turns a consumer web chat into a machine-readable API. That means models like GPT-4o-mini and Claude 3 Haiku become accessible from any codebase already speaking OpenAI’s JSON format, without rewriting your request logic.

Key highlights

• OpenAI-compatible chat completions endpoint with streaming support
• Backed by DuckDuckGo’s model menu: GPT-4o-mini, o3-mini, Claude 3 Haiku, Llama 3.3 70B, and Mixtral 8x7B (GPT-3.5 was dropped by DuckDuckGo)
• Built-in web UI served at /web
• Optional authorization, TLS termination, and proxy pooling via environment variables
• Ships as a single Go binary or a Docker image

Caveats

• DuckDuckGo officially removed GPT-3.5-Turbo support, so that model is gone for good
• The README is almost entirely deployment commands; architectural details and error-handling behavior are left unexplained

Verdict

A handy utility if you want a self-hosted bridge to free LLMs without managing API keys or billing. Look elsewhere if you need guaranteed uptime, official model support, or anything more sophisticated than a straight chat proxy.