When your stack speaks OpenAI but your contract is with Azure

What it does This Go service sits between OpenAI-speaking clients and Azure OpenAI, translating requests and responses so tools like ChatGPT-Next-Web, LangChain, and chatgpt-web work unmodified. It handles the deployment-specific URLs, model name remapping, and API version quirks Azure requires, so your existing code only needs a new base URL.

The interesting bit The real work is boring in the best way: mapping gpt-3.5-turbo to whatever arbitrary deployment name your Azure admin chose, and rewriting paths from /v1/chat/completions to Azure’s resource-scoped endpoints. It can even route different models to separate Azure endpoints and keys via a config file, which is handy when your embeddings live in East US and GPT-4 lives in West Europe.

Key highlights

• Verified compatibility with ChatGPT-Next-Web, chatgpt-web, Chatbox, and LangChain.
• Model name remapping (e.g., gpt-3.5-turbo → gpt-35-turbo) via environment variables or YAML config.
• Per-model endpoint and API key configuration in config.yaml for multi-region or multi-account setups.
• Outbound HTTP and SOCKS5 proxy support for corporate network egress.
• Ships as a lightweight Docker image.

Verdict Grab it if you are locked into Azure OpenAI but want to keep using the broader OpenAI ecosystem’s tools and SDKs. If you already call Azure’s REST API directly, this adds a hop you don’t need.