How to go further

This first tutorial aimed at priming the basic concepts. Some more advanced topics include:

• Streaming responses — real-time output for chat-like experiences (you dont have to wait for the end of the message to display it. Only useful for interactive sessions)
• Multi-step chains — feed one LLM’s output into another
• Tool calling (function calling) — let the LLM decide when to run your Python functions ( only run it if you are experienced and can setup protected environments)
• RAG from Python — query your own documents programmatically (i.e. let LLM scan documents via embedding models)
• Structured data extraction — parse LLM output into pandas DataFrames
• Local models via Ollama — same code but with local models instead of using OpenRouter (highly limited by the hardware/memory available)
• Rate limiting and retry logic — production-grade reliability: avoid draining your credits by accident

---