good first issue: docs/ollama.md local-server walkthrough

[primeodin]

Why

The README mentions OPENAI_BASE_URL for Ollama, but newcomers still stall on the exact env vars and a known-good smoke command.

Done when

• docs/ollama.md at repo root docs/
• Steps: install/run Ollama → pull a small model → set OPENAI_BASE_URL, OPENAI_API_KEY (any non-empty), OPENAI_MODEL → run one real (non-mock) prompt
• Notes what fails if the server is down (clear error path)
• Link from README "Real chat (optional)" section to this doc
• Keep it short (one screen)

Out of scope

Changing the client code or adding Docker.

Part of the PrimeOdin daily AI teaching builds.