- patients: move status filter to its own "Filter" row above the table
- patient sheet: name + ⋯ menu on one left row; Edit moved into the ⋯ menu
- settings: redesign sections with the COSS CardFrame surface (SettingsFrame)
- AI: add Automatic (auto-pick provider) and Off modes; default to Automatic
so a fresh install shows the setup banner until a provider is configured
- AI: fix the setup banner never showing (defaulted Ollama URL counted as
configured); add an "AI off" notice; add a fallback render for unknown chat
data parts so cards never silently vanish
- backend: include patient attachment metadata in the wallet record-update
bundle so pushed documents reach the wallet
- i18n: add mode/off/card keys across all five locales
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The chat defaulted to a Claude model, so a user who saved only a Gemini key got
a silent failure (backend derived Anthropic, found no key, errored — and the UI
showed nothing).
- Backend: resolveModel now falls back to whichever provider actually has a key
(preferring the configured one), so a Gemini key just works regardless of the
picked model. Clear 400 if no provider is configured at all.
- Frontend: the chat seeds its model/effort from the saved AI config, and now
surfaces request failures as both a persistent alert banner and a toast —
never silent.
- Settings: switching provider auto-selects that provider's default model.
- Refreshed the model catalog to current ids (Gemini 2.5 Pro/Flash + 2.0 Flash;
dropped the retired gemini-1.5-pro); per-provider default model updated.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Real LLM chat replacing the mock, backend-centric per the plan:
- Multi-provider API-key mode (OpenAI / Anthropic / Gemini via the AI SDK) plus
local Ollama (OpenAI-compatible endpoint). Provider is derived from the
picked model id; the matching stored key is used. New user_ai_settings table
holds per-user config with provider API keys encrypted at rest (AES-256-GCM,
src/lib/crypto.ts, keyed by AI_CREDENTIALS_KEY).
- POST /api/ai/config (get/put, secrets never returned), POST /api/ai/test
(Ollama ping / key presence), POST /api/ai/import (approved migration commit,
re-validated server-side, reuses the audited patient service).
- POST /api/chat: streamText agent with tools (getPatient, getPatientLabs,
searchPatients, previewImport). Real record data streams to the clinician as
custom data parts (cards) while the model sees only Veil-redacted results.
- Veil (src/services/ai/veil.ts): de-identifies patient identifiers to tokens
before external calls, resolves tokens on tool args, and rehydrates the final
answer. Bypassed for local Ollama. External mode runs non-streamed so the
rehydrated text is correct. Every call is audited (provider + Veil level).
- Shared role-scoping helpers extracted to src/lib/role-scope.ts (reused by the
patient routes and chat tools so visibility rules match).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Khalid Abdi