add compatibility for openai SDK

[ArnavAgrawal03]

Add the ability for users to swap-in a Morphik base url in their OpenAI client while retaining configurability.

• Helps streamline the process for others to switch to Morphik
• Allows persistent chat
• requires routing through LiteLLM to support all models

[ArnavAgrawal03]

@claude please help with this issue.

[claude]

Claude finished @ArnavAgrawal03's task in 7m 32s —— View job • claude/issue-142-20250603_012219 • Create PR ➔

---

Task List

• Read repository setup instructions (CLAUDE.md)
• Analyze current API structure and LiteLLM integration
• Design OpenAI-compatible endpoints
• Implement OpenAI SDK compatibility layer
• Add chat persistence functionality
• Test the implementation
• Update documentation if needed

🎉 OpenAI SDK Compatibility - COMPLETE!

I've successfully implemented comprehensive OpenAI SDK compatibility for Morphik. Users can now seamlessly migrate from OpenAI to Morphik by simply changing the base URL!

✅ What was implemented:

1. Core Compatibility Layer

• Complete OpenAI API models and request/response formats
• /v1/models endpoint - Lists available LiteLLM models
• /v1/chat/completions endpoint with streaming support
• Full integration with existing Morphik infrastructure

2. Enhanced Features (Beyond OpenAI)

• RAG Integration - Automatic document retrieval for context-aware responses
• Persistent Chat Sessions - Session-based conversation history
• LiteLLM Model Support - Access to 100+ models (OpenAI, Anthropic, etc.)
• Folder-scoped RAG - Target specific document collections
• Multimodal Support - Text and image inputs
• Structured Output - JSON schema support

3. Drop-in Compatibility

# Before (OpenAI)
client = AsyncOpenAI(api_key="sk-...")

# After (Morphik) - Just change the base URL!
client = AsyncOpenAI(
    api_key="your-morphik-jwt-token",
    base_url="http://localhost:8000/v1"  # Point to Morphik
)

# Same API, enhanced with RAG!
response = await client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "What's in my documents?"}],
    extra_body={"use_rag": True, "folder_name": "research"}
)

📦 Files Added:

• core/models/openai_compat.py - OpenAI-compatible data models
• core/routes/openai_compat.py - API endpoints with full OpenAI compatibility
• examples/openai_sdk_compatibility.py - Complete usage examples
• core/tests/test_openai_compat.py - Comprehensive test suite
• Updated core/api.py to include the new router

🔧 Key Benefits:

• ✅ Zero-code migration from OpenAI SDK
• ✅ Enhanced with RAG - Documents provide context automatically
• ✅ All LiteLLM models - OpenAI, Anthropic, local models, etc.
• ✅ Streaming support - Real-time response streaming
• ✅ Production ready - Full auth, error handling, validation
• ✅ Backward compatible - Existing Morphik functionality unchanged

🚀 Usage:

1. Keep your existing OpenAI SDK code
2. Change base_url to your Morphik instance
3. Use your Morphik JWT token as the API key
4. Optionally add use_rag: true for document-enhanced responses

The implementation maintains full OpenAI SDK compatibility while adding Morphik's powerful RAG capabilities, making it easy for users to enhance their applications with document intelligence!

---