Frequently Asked Questions (FAQ)

Q: What is the Gemini ➜ OpenAI API proxy? A: It’s a serverless proxy that allows you to use the free (within limits) Gemini API with tools and platforms designed for the OpenAI API.

Q: Why would I use this proxy? A: To leverage the Gemini API’s capabilities with tools that only support the OpenAI API, and to potentially save costs by using Gemini’s free tier.

Q: Is the Gemini API really free? A: Yes, but it has usage limits. Check the Gemini API documentation for details.

Q: What deployment options are available? A: You can deploy to Vercel, Netlify, Cloudflare Workers, Deno, or run it locally with Node, Deno, or Bun.

Q: How do I deploy to Vercel? A: You can use the “Deploy with Vercel” button or the Vercel CLI (vercel deploy).

Q: What is an API base and how do I configure it? A: The API base is the URL of your deployed proxy. It should be in the format https://your-proxy-url/v1. You need to configure your OpenAI-compatible tools to use this API base.

Q: What API endpoints are supported? A: Currently, chat/completions, embeddings, and models are supported. chat/completions has the most comprehensive parameter support.

Q: How do I use web search with Gemini through this proxy? A: Append :search to the model name (e.g., gemini-2.0-flash:search).

Q: What is UBOS and how does this proxy relate to it? A: UBOS is a full-stack AI Agent Development Platform. This proxy can be used to integrate Gemini into UBOS-based AI Agent workflows.

Q: Can I use vision and audio input with this proxy? A: Yes, vision and audio input are supported as per OpenAI specifications, implemented via inline data.

Q: What if I encounter a 404 Not Found error when opening my deployed site in a browser? A: This is expected. The API is not designed for direct browser access. You should use it through an OpenAI-compatible tool or application.