Health Platform v2 — Architecture Blueprint

• What it is: A system-design hub for rewriting Healthcare platform from a Django monolith into microservices. The folder ~/development/salomatic_v2 is NOT a git repo and contains docs/ (per-service architecture) + team/ + .claude/. It is the "map" and spec of the v2 platform; the implementation lives in separate service repositories (many of which are tasks of their own: phi-service #44, pii_salomatic #47, anon-service #7, lab-processor #26, salomatic_temporal #68, salomatic_lab_preview #66).
• Type / status / role: other (architecture / planning) · active · lead (architect) — the user designs the platform decomposition.
• Activity period: n/a (not git). Docs are dated Dec 2025 – Feb 2026 by mtime.

• Designed the full microservices architecture of v2: flow diagrams, services table (stack/port/purpose), per-service specifications.
• Defined the core privacy model: splitting PII/PHI/Anon across 3 services with anonymous mapping.
• Locked in infrastructure decisions (Temporal, MinIO, API Gateway).
• (Implementation is being done in separate service repos — see related tasks.)

Confirmed by docs/ARCHITECTURE.md:

• Privacy via three-way PII/PHI/Anon separation (pii_salomatic + phi-service + anon-service): personal data (PII) with field-level encryption and blind index (search over encrypted data), medical data (PHI) with two-level encryption, plus a dedicated anon-service mapping PII ↔ anonymous UUID. The services talk to each other, but none of them holds the full picture. → advanced privacy-by-design architecture under HIPAA/GDPR/PDPL.
• Blind index for search over encrypted PII — a non-trivial cryptographic pattern (search without decrypting). → deep understanding of applied cryptography.
• AI processing on LangGraph (lab-processor): agentic workflows for lab-result analysis as a separate service.
• Temporal orchestration of inter-service processes + MinIO for binary artifacts (PDF reports). → durable, fault-tolerant design.
• Polyglot microservices (Bun/Elysia + Python/FastAPI) with an API Gateway — deliberate stack choice per service.

• Sessions found: 1 directory (2 transcripts + index) — the architecture was worked out with Claude Code.
• What was done with AI: designing v2 architecture/specs (likely — generating and iterating per-service design docs with AI).
• AI-workflow patterns: AI-assisted system design (using AI to design a microservices architecture and its documentation).

• Full microservices architecture of a medical platform (5 core + support services).
• Privacy-by-design under 3 regulatory regimes (HIPAA/GDPR/Saudi PDPL).
• 8+ per-service specifications.