Services

Service	Port	Runtime	Purpose	Key specs
apps/web	6200	Next.js 15 · React 19 · Node 20	UI, patient + provider + admin consoles, decision-graph UI	Spec 014, 015, 017, 018, 056, 058
apps/api	6201	NestJS 10 · TypeScript strict · Node 20	Core business logic, envelope encryption, RBAC + consent, audit	Spec 001, 002, 004-015, 018, 019, 020, 021
apps/ml	6202	FastAPI · Python 3.12 · Pydantic v2	ASR, NER, CDS, Coding AI with dual-LLM	Spec 001 (FR-006), 041
memory-store	6401	FastAPI · motor · Python 3.12	Knowledge graph + hybrid retrieval + MCP server	Specs 026, 028, 042-047, 060
research-engine	6402	FastAPI · motor · Python 3.12	5 agents + 12 trusted-source connectors + change-stream pipeline	Specs 032, 048, 049, 050, 051, 055, 061
orchestrator	6403	FastAPI · motor · Python 3.12	6-phase hypothesis candidate lifecycle	Specs 027, 034, 037, 040, 052, 053, 059
model-optimization	6404	FastAPI · motor · Python 3.12	Decision graph + eval + outcome feedback + benchmark	Specs 030, 033, 035, 038, 057, 062
apps/landing	static	Plain HTML + Tailwind CDN	Public marketing page (deployed to Azure SWA)	—
apps/docs	static	Docusaurus	This site (deployed to Azure SWA)	—

Inter-service comms

All comms are HTTP + service-account JWT (FR-KIA-012). No shared DB connections, no cross-service Python imports.

Database footprint

Service	Collections accessed
apps/api	users, patients, encounters, notes, care_teams, referrals, consents, audit_events, organizations, family_groups, notifications, ehr_imports, refresh_sessions
apps/ml	llm_call_logs (writes encrypted prompt/output; posts audit to api via service JWT)
memory-store	memory_entities, memory_relationships, memory_claims, research_sources, retrieval_audit
research-engine	research_tasks, correlator_findings, replicator_findings, librarian_findings (+ reads memory-store via HTTP)
orchestrator	orchestration_candidates, orchestration_runs, candidate_outcomes (writes only; reads memory-store via HTTP)
model-optimization	experiment_traces, decision_graph_nodes, dpo_preference_pairs (reads llm_call_logs.metadata + retrieval_audit)

Every PHI-bearing write goes through FieldEncryptionService (FR-005).

Health checks + observability

Service	Health endpoint	Audit source
apps/api	GET /api/health	api
apps/ml	GET /api/ml/health	ml
memory-store	GET /api/memory/health	memory-store
research-engine	GET /api/research/health	research-engine
orchestrator	GET /api/orchestrator/health	orchestrator
model-optimization	GET /api/optimization/health	model-optimization

Every service exports:

• Azure Monitor traces (distributed via W3C trace-context)
• OpenTelemetry counters on per-route latency + status
• Audit events (every security-relevant action)