[assess] Agentic guard against junk/jailbreak input; tidy economic UX

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

[fix] Treat chatbot-greeting assessments as weak (use grounded fallback)

[fix] Robust assessment + 2-3 recs + documents progress

[assess] Remove web search entirely (prompt, code, module, test, env)

[fix] Hide doubled 'processing' spinner; never leak @@END / metadata

[fix] Choice controls never toggle visibility; non-signatories not asylum dests

[assess] Honest handling of economic (non-protection) cases

[ux] Show multi-select options instantly (no extra Continue)

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

[fix] Fully-scripted multilingual interview (no LLM) + remove walkthrough chrome

[fix] Deterministic slot-filling interview (agentic flow, consistent UI)

[T049-T055 + UX] Phase 5 documents + single-model + responder/UX fixes

[T044-T048 + data] Phase 4 recommendations + real UNHCR stats integration

[T035-T043] Phase 3 — situation assessment (reasoning stream + real tools)

[T030-T034] Phase 2 — intake screen + language selection + theming

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

[T011+] Env config: uv venv, .env loader, remote Ollama host + model

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