Feature/langgraph gemini example

[ParthSingh0506]

Summary

Adds a complete LangGraph + Google Gemini integration example demonstrating APO (Automatic Prompt Optimization) training.

Features

• LangGraph integration with Google Gemini 2.5 Flash (non-OpenAI LLM)
• Bulletproof placeholder handling to prevent KeyErrors from APO's creative prompt generation
• Persistent training result storage (JSON, TXT, PKL formats)
• Windows compatibility fixes for path handling
• Comprehensive README with installation, troubleshooting, and customization guides

What's Included

• agent_train.py - Complete training script with news agent + APO setup
• README.md - Detailed setup instructions and troubleshooting
• requirements.txt - All Python dependencies
• .env.example - API key configuration template
• .gitignore - Prevents accidental secret commits

Problem Solved

This example addresses common issues developers face when integrating LangGraph with Agent Lightning:

1. KeyError on unexpected placeholders - APO's GPT-4 editor creates prompts with placeholders beyond {query}, causing runtime errors
2. Infinite training loops - Missing beam_rounds parameter causes training to run indefinitely
3. Lost optimization results - Incorrect prompt extraction methods lose the optimized prompt

Testing

Tested on:

• ✅ Windows 11 with Python 3.12
• ✅ Standard pip + venv workflow
• ✅ Successfully completes 3 training rounds
• ✅ Produces optimized prompt with improved validation scores

Related

This example fills a gap in the current examples - there are no LangGraph integrations or non-OpenAI LLM examples in the repository yet.

Note: I'm writing a companion blog article with a detailed walkthrough that
can serve as additional community documentation.

[ParthSingh0506]

@microsoft-github-policy-service agree

[ultmaster]

Hi. Thanks a lot for the contribution!

Would you be willing to move the example to contrib/recipes. That's where we put community examples that have not been tested on CI yet. We can graduate it when we have time.

[ParthSingh0506]

Hi @ultmaster Thanks for the quick feedback! I'd be happy to move the example to contrib/recipes.

I noticed the test failures are in the openai_agents_sdk_mcp_tool_use test, which appears to be unrelated to my LangGraph example. The pre-commit hooks also made some formatting changes to my code.

Should I:

1. Apply the pre-commit formatting changes locally and push them?
2. Move the files to contrib/recipes in the same PR?

Or would you prefer I close this and open a new PR directly in contrib/recipes with the formatting fixes?

Happy to do whichever is easiest for you!

[ultmaster]

1. Yes
2. In the same PR

The openai_agents_sdk_mcp_tool_use should pass now. Please merge from main.

[ParthSingh0506]

Done! I've merged the latest changes from main, applied all pre-commit formatting fixes, and moved the example to contrib/recipes/langgraph-news-agent as requested.

Ready for review!

[ParthSingh0506]

Hi @ultmaster ,

Just following up on this! The changes are ready and I've moved everything to contrib/recipes as you requested. Would you be able to approve the workflow so the CI checks can run?

Happy to fix any issues that come up. Thanks for your time!

[ParthSingh0506]

@ultmaster Hi! Checking in on this PR again.

I completed the requested changes a couple weeks ago:

• Merged from main
• Applied pre-commit formatting
• Moved to contrib/recipes/langgraph-news-agent

The workflow just needs approval to run the CI checks. I'm ready to quickly fix any test failures that come up.

Would really appreciate if you could approve the workflow when you get a chance, or let me know if there's anything else needed!

Thanks!

[ultmaster]

hi. was on vacation and just back. let's check the CI.

[ParthSingh0506]

Hi @ultmaster Thanks for approving the workflow!

I've added the missing copyright header to agent_train.py and pushed the fix.

I'm also seeing test failures in the Lint - next job, but those appear to be in the core framework files (agentlightning/instrumentation/weave.py and agentlightning/tracer/weave.py) rather than my contribution. The errors are related to missing type definitions for OtelExportReq and OtelExportRes.

My contribution files in contrib/recipes/langgraph-news-agent/ should now pass all checks. Let me know if there's anything else I need to address!