OpenAI-compatible API gateway for ChatGPT, DeepSeek, Qwen, Claude & Gemini using browser automation

A unified OpenAI-compatible API gateway for multiple AI chat services including ChatGPT, DeepSeek, Qwen, Claude, and Gemini.

This solution enables seamless integration with popular AI platforms through a standardized REST API interface, leveraging browser automation to interact with web-based chat interfaces.

This project provides local API servers that expose OpenAI-compatible endpoints for various AI chat services. By utilizing Playwright for browser automation, it captures real-time streaming responses and delivers them through a familiar API format.

• OpenAI-Compatible API - Drop-in replacement for OpenAI SDK integrations
• Multi-Provider Support - ChatGPT, DeepSeek, Qwen, Claude, and Gemini in one solution
• Real-Time Streaming - Server-Sent Events (SSE) support for streaming responses
• Image Generation - Nano Banana (Gemini) and Qwen image generation support
• Concurrent Request Handling - Dynamic page pool supporting up to 10 parallel requests
• Session Persistence - Automatic browser state management for seamless authentication
• Stateless Architecture - Each request initiates a fresh conversation

┌──────────────────┐     ┌──────────────────┐     ┌──────────────────┐
│   Application    │────▶│   API Gateway    │────▶│  Browser Engine  │
│  (Your Client)   │     │   (Node.js)      │     │   (Playwright)   │
└──────────────────┘     └──────────────────┘     └──────────────────┘
                                                           │
                                                           ▼
                                                  ┌──────────────────┐
                                                  │   AI Platform    │
                                                  │ (ChatGPT, etc.)  │
                                                  └──────────────────┘

The gateway intercepts network traffic to capture streaming responses directly, ensuring accurate and efficient data transmission.

├── chatgpt/               # ChatGPT Gateway (Port 3003)
│   ├── server.js
│   ├── package.json
│   └── browser-state.json
│
├── deepseek/              # DeepSeek Gateway (Port 3000)
│   ├── server.js
│   ├── package.json
│   └── browser-state.json
│
├── qwen/                  # Qwen Gateway (Port 3001)
│   ├── server.js
│   ├── package.json
│   └── browser-state.json
│
├── claude/                # Claude Gateway (Port 3002)
│   ├── server.js
│   ├── package.json
│   └── browser-state.json
│
├── gemini/                # Gemini Gateway (Port 3004)
│   ├── server.js
│   ├── package.json
│   └── browser-state.json
│
└── README.md

• Node.js 18.x or higher
• Chromium browser (automatically installed by Playwright)

Navigate to the desired provider directory and install dependencies:

cd chatgpt  # or deepseek, qwen, claude, gemini
npm install

node server.js

Note: On first launch, a browser window will open for authentication. Complete the login process manually. Session credentials are automatically persisted for subsequent runs.

ChatGPT

DeepSeek

Qwen

Claude

Gemini

Gemini Image Generation: Use nano-banana or nano-banana-pro models to generate images. The response includes base64-encoded image data that can be displayed directly.

The Gemini gateway includes a built-in chat interface for testing:

1. Start the Gemini server: cd gemini && node server.js
2. Open gemini/chat.html in your browser
3. Features:
   • Model selection (all 5 models)
   • Streaming toggle
   • Markdown rendering
   • Image display for Nano Banana responses
   • Connection status indicator

from openai import OpenAI

# Initialize client
client = OpenAI(
    base_url="http://localhost:3003/v1",
    api_key="sk-chatgpt"
)

# Standard completion
response = client.chat.completions.create(
    model="gpt-5.1",
    messages=[
        {"role": "user", "content": "Explain quantum computing"}
    ]
)
print(response.choices[0].message.content)

# Streaming completion
stream = client.chat.completions.create(
    model="gpt-5.1",
    messages=[{"role": "user", "content": "Write a short story"}],
    stream=True
)
for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

# Standard request
curl -X POST http://localhost:3003/v1/chat/completions \
  -H "Authorization: Bearer sk-chatgpt" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.1",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": false
  }'

# Streaming request
curl -X POST http://localhost:3003/v1/chat/completions \
  -H "Authorization: Bearer sk-chatgpt" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.1",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'

const response = await fetch("http://localhost:3003/v1/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": "Bearer sk-chatgpt",
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: "gpt-5.1",
    messages: [{ role: "user", content: "Hello" }],
    stream: false
  })
});

const data = await response.json();
console.log(data.choices[0].message.content);

$headers = @{
    "Authorization" = "Bearer sk-chatgpt"
    "Content-Type" = "application/json"
}

$body = @{
    model = "gpt-5.1"
    messages = @(
        @{ role = "user"; content = "Hello" }
    )
    stream = $false
} | ConvertTo-Json

$response = Invoke-RestMethod -Uri "http://localhost:3003/v1/chat/completions" `
    -Method POST -Headers $headers -Body $body

$response.choices[0].message.content

# Text generation with Gemini Preview
curl -X POST http://localhost:3004/v1/chat/completions \
  -H "Authorization: Bearer sk-gemini" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-preview",
    "messages": [{"role": "user", "content": "Explain quantum computing"}]
  }'

# Image generation with Nano Banana
curl -X POST http://localhost:3004/v1/images/generations \
  -H "Authorization: Bearer sk-gemini" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana",
    "prompt": "A cat astronaut floating in space"
  }'

# Image generation via chat endpoint
curl -X POST http://localhost:3004/v1/chat/completions \
  -H "Authorization: Bearer sk-gemini" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana-pro",
    "messages": [{"role": "user", "content": "Generate an image of a sunset over mountains"}]
  }'

# PowerShell - Gemini text request
$response = Invoke-RestMethod -Uri "http://localhost:3004/v1/chat/completions" `
  -Method POST `
  -Headers @{"Authorization"="Bearer sk-gemini"; "Content-Type"="application/json"} `
  -Body '{"model":"gemini-preview","messages":[{"role":"user","content":"Hello"}]}'

$response.choices[0].message.content

PORT=8080 HEADLESS=true MAX_PAGES=20 node server.js

The gateway implements a dynamic page pool for handling concurrent requests:

• Requests are assigned to available browser pages
• New pages are created on-demand up to MAX_PAGES limit
• Idle pages are automatically recycled after 60 seconds
• Minimum of 1 page is always maintained

• API keys are local identifiers, not transmitted externally
• Browser sessions contain authentication tokens - secure browser-state.json appropriately
• Intended for local development and personal use

This project is provided for educational and personal use. Users are responsible for compliance with respective AI provider terms of service.