Gemini 2 OpenAI Proxy

This program is a wrapper for Gemini CLI that can serve Google Gemini 2.5 Pro (or Flash) through an OpenAI-compatible API. It works plug-and-play with clients that already speak OpenAI like SillyTavern, llama.cpp, LangChain, VS Code's Cline extension, etc.

---

Features

✔	Feature	Notes
/v1/chat/completions	Non-stream & stream (SSE)	Works with curl, ST, LangChain…
Vision support	image_url → Gemini inlineData
Function / Tool calling	OpenAI "functions" → Gemini Tool Registry
Reasoning / chain-of-thought	Sends enable_thoughts:true, streams <think> chunks	ST shows gray bubbles
1M token context	Proxy auto-elevates Gemini CLI's default 200k limit
CORS	Enabled (*) by default	Ready for browser apps

---

Quick Start

With npm

git clone https://github.com/DouveAlexandre/GeminiCLI-Proxy2OpenAIAPI
cd GeminiCLI-Proxy2OpenAIAPI
npm i
npm start # starts the server (runs on port 11434 by default)

With Docker

Alternatively, you can use the provided Dockerfile to build a Docker image.

docker build --tag "GeminiCLI-Proxy2OpenAIAPI" .
docker run -p 11434:80 -e GEMINI_API_KEY=your_key_here GeminiCLI-Proxy2OpenAIAPI

---

.env File Configuration

The application uses an .env file for configuration. Create an .env file in the project root with the following variables:

Basic Configuration

# Port where the server will run
PORT=11434

# Authentication type - can be 'oauth-personal', 'gemini-api-key', or 'vertex-ai'
AUTH_TYPE=oauth-personal

# Path to OAuth credentials (only for AUTH_TYPE='oauth-personal')
GOOGLE_APPLICATION_CREDENTIALS=C:\Users\your_user\.gemini\oauth_creds.json

# Gemini API Key (only for AUTH_TYPE='gemini-api-key')
GEMINI_API_KEY=

# Model to be used - 'gemini-2.5-flash' or 'gemini-2.5-pro'
# Leave empty to use CLI default model
MODEL=

Authentication Types

1. OAuth Personal (Recommended for personal use)

AUTH_TYPE=oauth-personal
GOOGLE_APPLICATION_CREDENTIALS=C:\Users\your_user\.gemini\oauth_creds.json

• Advantages: Free access to Gemini 2.5 Pro
• How to configure: Log in to your Google account through Gemini CLI
• Requirements: Gemini CLI configured with OAuth

2. Gemini API Key

AUTH_TYPE=gemini-api-key
GEMINI_API_KEY=your_api_key_here

• Advantages: Simpler to configure
• How to obtain: Visit Google AI Studio
• Requirements: Valid Google AI API key

3. Vertex AI

AUTH_TYPE=vertex-ai

• Usage: For corporate environments with Vertex AI
• Requirements: Additional Google Cloud configuration

Model Configuration

# To use Gemini 2.5 Flash (faster, less accurate)
MODEL=gemini-2.5-flash

# To use Gemini 2.5 Pro (slower, more accurate)
MODEL=gemini-2.5-pro

# Leave empty to use default model
MODEL=

---

How to Use

1. Initial Setup

1. Clone the repository and install dependencies:

git clone https://github.com/DouveAlexandre/GeminiCLI-Proxy2OpenAIAPI
cd GeminiCLI-Proxy2OpenAIAPI
npm install

2. Create the .env file with your settings (see section above)

3. Start the server:

npm start

The server will be available at http://localhost:11434

2. Test with curl

curl -X POST http://localhost:11434/v1/chat/completions \
     -H "Content-Type: application/json" \
     -d '{
       "model": "gemini-2.5-pro",
       "messages": [{"role": "user", "content": "Hello Gemini!"}]
     }'

3. SillyTavern Configuration

1. Go to API Connections
2. Select Chat Completion (OpenAI, Claude, Window/OpenRouter, Scale)
3. Configure:
   • API: OpenAI
   • API Base URL: http://127.0.0.1:11434/v1/chat/completions
   • API Key: any value (not used)
   • Model: gemini-2.5-pro or gemini-2.5-flash

4. VS Code Configuration (Cline)

1. Install the Cline extension
2. Configure:
   • Provider: OpenAI Compatible
   • Base URL: http://localhost:11434/v1/chat/completions
   • API Key: any value
   • Model: gemini-2.5-pro

5. Using with LangChain (Python)

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="http://localhost:11434/v1/chat/completions",
    api_key="any-value",
    model="gemini-2.5-pro-latest"
)

response = llm.invoke("Hello, how are you?")
print(response.content)

---

Advanced Features

Image Support

curl -X POST http://localhost:11434/v1/chat/completions \
     -H "Content-Type: application/json" \
     -d '{
       "model": "gemini-2.5-pro-latest",
       "messages": [{
         "role": "user",
         "content": [
           {"type": "text", "text": "Describe this image"},
           {"type": "image_url", "image_url": {"url": "data:image/jpeg;base64,..."}}
         ]
       }]
     }'

Streaming

curl -X POST http://localhost:11434/v1/chat/completions \
     -H "Content-Type: application/json" \
     -d '{
       "model": "gemini-2.5-pro-latest",
       "messages": [{"role": "user", "content": "Tell me a story"}],
       "stream": true
     }'

Function Calling

curl -X POST http://localhost:11434/v1/chat/completions \
     -H "Content-Type: application/json" \
     -d '{
       "model": "gemini-2.5-pro-latest",
       "messages": [{"role": "user", "content": "What's the weather today?"}],
       "tools": [{
         "type": "function",
         "function": {
           "name": "get_weather",
           "description": "Get weather information",
           "parameters": {
             "type": "object",
             "properties": {
               "location": {"type": "string", "description": "Location"}
             }
           }
         }
       }]
     }'

---

Troubleshooting

OAuth Authentication Error

• Check if the path in GOOGLE_APPLICATION_CREDENTIALS is correct
• Run gemini auth in the terminal to reauthenticate

API Key Error

• Check if the key in GEMINI_API_KEY is valid
• Confirm that AUTH_TYPE=gemini-api-key

Server does not start

• Check if the port specified in PORT is available
• Confirm that all dependencies were installed with npm install

Model not found

• Use gemini-2.5-pro or gemini-2.5-flash as the model name
• Leave MODEL empty in .env to use the default

---

License

MIT – free for personal and commercial use. Fork of https://huggingface.co/engineofperplexity/gemini-openai-proxy