[M5] Local Model Support (Ollama)

[cursor]

Milestone 5 — Embedded AI via Embabel | Priority: Medium | PARTIALLY DONE in PR #27

OllamaProvider for local LLM text generation.

• OllamaProvider implemented in ai_provider.py (chat via REST API)
• HybridProvider routes chat→Ollama, TTS/transcribe→OpenAI
• Remaining: streaming support, auto-detect Ollama availability, documentation

See issues/embedded-ai-embabel/18-local-model-support.md for full spec.