[fix] Give Ollama a real context window (root cause); deterministic weak-check

[deploy] Modal-hosted Ollama backend + remote auth headers

[agentic] Port pi loop hooks + LLM-drafted editable (Word) documents

[T017-T029] Phase 1 — agent loop + interview core

[T006] Loop: gracefully degrade when model rejects `think` param

[T001-T016] Phase 0 — project setup: agent loop, state machine, Gradio skeleton