Handle OpenRouter retries and reset flows

Retire v1 template and refine analysis config UI

Add care experience breakdown pie chart

Add rubric instructions and attributes to frameworks

Analysis frameworks: themes, top-down instructions, category descriptions

Analysis: unify templates into frameworks

Feat: split analysis into 3 sequential passes

UI: select analysis template per run

Fix: robust OpenRouter parsing and hide raw errors

Fix analysis template seeding (FK order)

Analysis templates: lazy-seed default template

Analysis templates: dynamic top-down codebook

Analysis agent: add system prompt setting with UI warning

Make style guidance user-visible via system prompts

DB-canonical personas: CRUD + shared settings

Feat: universal persona system prompts + attributes-only defaults

Chore: align analysis defaults with prior rules

Feat: analysis agent attributes + defaults

Feat: patient attributes v1 (local-first)

Fix: persist runs by passing asked_question_ids

UX: hide ids + human turn gating + save flash

UI: move persona selectors into panels + add AI role toggle

UI: align surveyor config + remove prompt addition

Refactor: replace surveyor knobs with attributes list

Feat: question bank JSON selection

Plumb: surveyor question bank config

Feat: surveyor core knobs (compiled prompt + UI)

Fix: support older aiosqlite without execute_fetch*

Fix: stop conversation when last WS client disconnects

Feat: persist sealed runs to SQLite

UI: rename chat tabs

Refactor: remove legacy catch-all routes

Refactor: modularize API routers and UI views

Upload: analyze pasted text and PDFs

Export: download Excel + JSON

Fix HF login by using token auth fallback

Add config view and shared-password gate

Fix ConversationInfo dataclass field order

Config panel MVP: persona selection + prompt additions

Reduce verbosity: role constraints + token limits

Top-down codebook: add categories output and UI

Bottom-up findings: add short code label

Normalize resource-agent confidence to 0..1

Resource agent: add mixed care box and avoid duplicated evidence

Resource agent: add neutral/additional care box

Resource agent: split care experience into positive/negative

Resource agent: cite evidence via sentence IDs (pysbd)

Cleanup completed conversations after resource analysis

MVP resource agent: analyze transcript on completion

Fix backend imports for Docker (core -> backend.core)