universal_proxy_demo

Universal API Proxy Demo

This example demonstrates how the universal API proxy can accept different provider API formats and route them to any backend provider.

What This Demo Shows

The same backend provider (e.g., Gemini) can respond to requests in three different API formats:

1. OpenAI format - /v1/chat/completions
2. Gemini format - /v1beta/models/{model}:generateContent
3. Anthropic format - /v1/messages

This means you can use any SDK or tool designed for one provider with a completely different backend provider!

Prerequisites

1. Start the universal API proxy server with your chosen backend provider:

# Example: Use Gemini as the backend for all three formats
export GEMINI_API_KEY="your-gemini-api-key"

# Terminal 1: OpenAI format -> Gemini backend
api-proxy -listen :8081 -format openai -provider gemini -model gemini-2.5-flash

# Terminal 2: Gemini format -> Gemini backend
api-proxy -listen :8082 -format gemini -provider gemini -model gemini-2.5-flash

# Terminal 3: Anthropic format -> Gemini backend
api-proxy -listen :8083 -format anthropic -provider gemini -model gemini-2.5-flash

Or run a single instance that accepts one format:

# Single instance accepting OpenAI format, routing to Gemini
export GEMINI_API_KEY="your-gemini-api-key"
api-proxy -listen :8080 -format openai -provider gemini -model gemini-2.5-flash

Running the Demo

# Make sure the proxy is running first!
go run main.go

Expected Output

================================================================================
Universal API Proxy Demo
================================================================================

Example 1: Using OpenAI Format
--------------------------------------------------------------------------------
Request URL: http://localhost:8080/v1/chat/completions
Request body: {"model":"gpt-4","messages":[{"content":"Say 'Hello from OpenAI format!'","role":"user"}]}

Response:
{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "created": 1234567890,
  "model": "gpt-4",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello from OpenAI format!"
      },
      "finish_reason": "stop"
    }
  ],
  ...
}

✓ Assistant's message: Hello from OpenAI format!

Example 2: Using Gemini Format
--------------------------------------------------------------------------------
Request URL: http://localhost:8080/v1beta/models/gemini-2.5-flash:generateContent
Request body: {"contents":[{"parts":[{"text":"Say 'Hello from Gemini format!'"}]}]}

Response:
{
  "candidates": [
    {
      "content": {
        "parts": [
          {
            "text": "Hello from Gemini format!"
          }
        ],
        "role": "model"
      }
    }
  ]
}

✓ Assistant's message: Hello from Gemini format!

Example 3: Using Anthropic Format
--------------------------------------------------------------------------------
Request URL: http://localhost:8080/v1/messages
Request body: {"max_tokens":1024,"messages":[{"content":"Say 'Hello from Anthropic format!'","role":"user"}],"model":"claude-3-5-sonnet-20241022"}

Response:
{
  "id": "msg_...",
  "type": "message",
  "role": "assistant",
  "model": "claude-3-5-sonnet-20241022",
  "content": [
    {
      "type": "text",
      "text": "Hello from Anthropic format!"
    }
  ],
  "stop_reason": "end_turn"
}

✓ Assistant's message: Hello from Anthropic format!

================================================================================
Demo Complete!

Note: All three requests can be routed to the SAME backend provider,
regardless of which API format they use. This is the power of the
universal proxy!
================================================================================

Use Cases Demonstrated

1. Format Flexibility: The same backend (Gemini) responds correctly to three different API formats
2. SDK Compatibility: Each format can be used with its corresponding SDK (OpenAI SDK, Gemini SDK, Anthropic SDK)
3. Migration Path: Easily migrate between providers without changing client code
4. Cost Optimization: Use cheaper providers while keeping your existing API format

Try Different Combinations

Change the proxy configuration to try different combinations:

# Use OpenAI format to call Anthropic
export ANTHROPIC_API_KEY="your-key"
api-proxy -format openai -provider anthropic -model claude-3-5-haiku-20241022

# Use Anthropic format to call OpenAI
export OPENAI_API_KEY="your-key"
api-proxy -format anthropic -provider openai -model gpt-4o-mini

# Use Gemini format to call OpenAI
export OPENAI_API_KEY="your-key"
api-proxy -format gemini -provider openai -model gpt-4o

All 9 combinations work!