fix(auth): allow unauthenticated access to /generate-report

feat(config): load .env on startup and document DEEPSEEK_API_KEY

feat(llm): switch report generator from Anthropic to DeepSeek

fix(counterfactuals): use disease-specific inference threshold instead of hardcoded 0.5

fix(nlp): extract age and BP from short/informal clinical notes

fix(api): add prediction and confidence to /explain response

fix(model): correct preprocessing order and invert heart disease labels

fix(auth): replace passlib with direct bcrypt to fix Python 3.10 compat

fix(startup): create DB tables and seed default users on first boot

fix(backend): patch retrain pipeline, admin error handling, PatientVisit import

fix(counterfactuals): enable What-If scenarios for heart_disease

feat: complete remaining roadmap features (1.3, 1.5, 1.6, 3.1, 4.3)

feat(nlp): wire Clinical Notes Parser into EMR mode

feat(admin): add Create User, Edit Role, Activate/Deactivate user management

fix(deploy): download preprocessors in Dockerfile + harden predict/counterfactuals

fix(startup): bake model into Docker image at build time

fix: add Optional to typing imports in main.py

fix(startup): start uvicorn immediately, download model in background lifespan

fix(auth): add missing get_optional_user dependency

fix(auth): allow unauthenticated access to predict/explain/counterfactuals

chore: remove stroke and CKD modules (no trained models available)

fix(diseases): add has_model check so stroke/CKD hidden until models trained

fix(startup): make new feature imports safe with try/except fallbacks

feat(federated): add per-hospital Flower client with XGBoost incremental training

feat(federated): add FedAvg server aggregator with differential privacy helper

feat(federated): add federated learning package init

feat(active-learning): wire review router + auto-queue uncertain predictions

feat(active-learning): add review queue API (list/annotate/skip/stats)

feat(active-learning): add entropy-based uncertainty sampler

feat(active-learning): add active_learning package init

feat(timeseries): add POST + GET /patients/{id}/visits endpoints

feat(timeseries): add visits relationship to Patient model

feat(timeseries): add PatientVisit SQLAlchemy model for longitudinal risk tracking

feat(nlp): add POST /api/v4/parse-notes endpoint

feat(nlp): add ClinicalNotesParser with BioBERT NER + regex fallback

feat(nlp): add nlp package init

feat(llm): add POST /api/v4/generate-report endpoint with rate limiting

feat(llm): add ClinicalReportGenerator with Anthropic API + rule-based fallback

feat(llm): add llm package init

feat(disease): add StrokeInput + CKDInput schemas and register in DISEASE_SCHEMA_REGISTRY

feat(security): add SecurityHeadersMiddleware with OWASP + HIPAA headers

feat(monitoring): wire monitoring router, Prometheus /metrics, and security headers

feat(monitoring): add admin drift + MLflow + Prometheus HTTP endpoints

feat(mlflow): add MLflow experiment tracker for model versioning

feat(metrics): add Prometheus metrics module for OmniDiag observability

feat(drift): add DriftMonitor using Evidently for data drift detection

feat(batch): add POST /api/v4/{disease}/batch endpoint for CSV predictions

feat(admin): add /admin/stats and /admin/users endpoints

refactor(backend): update core inference modules for v2 compatibility

feat(main): wire auth, RBAC, cache, rate-limit & error envelopes into FastAPI app