|
|
--- |
|
|
title: Vertex to Gemini Proxy |
|
|
emoji: 🚀 |
|
|
colorFrom: blue |
|
|
colorTo: green |
|
|
sdk: docker |
|
|
app_port: 7860 |
|
|
secrets: |
|
|
- PROXY_API_KEY |
|
|
- VERTEX_EXPRESS_KEYS |
|
|
--- |
|
|
|
|
|
# Vertex to Gemini Proxy |
|
|
|
|
|
This Hugging Face Space hosts a FastAPI application that acts as a proxy between a Vertex AI Express endpoint and the Gemini API. |
|
|
|
|
|
## Features |
|
|
|
|
|
- **Authentication**: Protects the proxy with an API key. |
|
|
- **Key Rotation**: Rotates through a list of Vertex Express keys. |
|
|
- **Project ID Extraction**: Automatically determines the Google Cloud Project ID from the Vertex Express key. |
|
|
- **Dynamic Proxy**: Forwards requests to the appropriate Gemini model and function. |
|
|
- **Streaming Support**: Handles streaming responses from the Gemini API. |
|
|
- **Model-Specific Logic**: Modifies request bodies for specific models as needed. |
|
|
|
|
|
## Usage |
|
|
|
|
|
1. Set the `PROXY_API_KEY` and `VERTEX_EXPRESS_KEYS` secrets in your Hugging Face Space settings. |
|
|
2. Make requests to the Space URL, following the Gemini API format. |
|
|
3. Provide the `PROXY_API_KEY` in the `x-goog-api-key` header or as a `key` query parameter. |
