[Enhancement] LiteLLM Integration

[psriramsnc]

🚀 SyGra works with LiteLLM!

Bringing more flexibility, more model support, and smoother integrations 🎉🤖

🧩 Summary

This PR integrates LiteLLM into SyGra, enabling seamless use of OpenAI, Azure OpenAI, and VLLM models through a unified backend.
It adds shiny new model wrappers ✨, updates the model factory ⚙️, introduces new tests 🧪, improves documentation 📚, and cleans up configuration logic 🧹.

⚙️ Features Implemented

• ✨ Added LiteLLM-backed model implementations (OpenAI, Azure OpenAI, VLLM)
• 🔧 Updated model factory to support backend: litellm
• 📦 Added LiteLLM dependency in pyproject.toml
• 🧪 Added comprehensive unit tests for LiteLLM-backed models
• 📚 Updated documentation for model configuration
• 🧹 Cleanup of imports, validation logic, and config parsing

🚄 Performance Impact

• 📉 No direct performance metrics included
• 🌐 LiteLLM may add slight routing overhead depending on setup
• 🟰 Overall: N/A — no measurable performance impact reported

🧪 How to Test

To verify the functionality end-to-end:

1. ▶️ Run any SyGra flow or LLM node using models with: openai, azure_openai, vllm.
   LiteLLM is now the default backend for these models. Ensure requests route through LiteLLM and outputs look correct.
2. 🔄 Test backward compatibility by setting backend: default in models.yaml.
   The pipeline should use the original custom models, not LiteLLM.
3. 🧪 Run make test-verbose to validate all new test cases are passing successfully.

📸 Screenshots (if applicable)

N/A 🖼️

✅ Checklist

• 🧹 Lint fixes and unit testing done
• 🔁 End-to-end task testing
• 📝 Documentation updated

🗒️ Notes

🎯 No behavioral changes for end users.
This update is fully backward compatible, and all existing flows will continue to work exactly as before.

[amitsnow]

LGTM 🚀

[zephyrzilla]

LGTM!

[amitsnow]

LGTM 🚀