spike: install Ollama + FunctionGemma local inference stack

[iAmGiG]

Summary

Set up local LLM inference infrastructure for FunctionGemma experimentation.

Background

FunctionGemma is a 270M parameter model purpose-built for function calling. Runs locally via Ollama with ~300MB footprint - suitable for laptop deployment.

Reference: https://blog.google/technology/developers/functiongemma/

Tasks

• Install Ollama on dev machine
• Pull FunctionGemma model (ollama pull functiongemma)
• Add autogen-ext[ollama] to requirements.txt
• Verify Ollama server runs on default port (11434)
• Basic smoke test: call FunctionGemma with simple prompt

Acceptance Criteria

• Ollama running locally
• FunctionGemma responds to basic function-calling prompts
• No changes to existing codebase yet

Notes

This is infrastructure setup only. No integration with existing code.