Sog'liqni saqlash uchun AI-native Work OS

• Bu nima: "bitta mahsulot, bitta repo" — MISS US healthtech platform'ning yagona mahsuloti sifatida joylashtirilgan bo'lib, turli xil mijoz turlariga alohida mahsulotlar sifatida emas, balki turli RBAC + konfiguratsiyaga ega Organizatsiyalar sifatida xizmat qiladi: klinik (De Factum — LIS+CIS+CRM+chartlar+BI), akkreditatsiya (IAC/CBAHI/SASE + Cortex), bemor ilovasi va US healthtech platformning o'zi (birinchi tenant, dogfooding). Jamoaning boshqa barcha repozitoriylari (healthcare-crm, regain-accreditation, regain-brain/Hermes/Deutsch/Popper/Bench, command-center) "singdirish" rejimida → MISS modullariga aylanmoqda.
• Turi / holati / roli: web-app (ulkan multi-tenant monorepo) / faol / contributor (asosiy — 2-muallif). Lider/asoschi — Anton Kim (US healthtech platform) ~4378 kommit; foydalanuvchi (Davron+Dave93) — ~760 kommit (~17%, 2-o'rin); shuningdek E2E botlar, Codex bot va boshqa hissa qo'shuvchilar.
• Faollik oynasi: 2026-02-09 → 2026-05-26 — ~3.5 oyda 4556 kommit (nihoyatda yuqori AI-driven tezlik).

• Monorepo: Turborepo + Bun; apps/ — web, admin, api (Elysia.js), queue (background/OCR), sites. Next.js 16 + React, Better Auth + chuqur RBAC (personalar: registered nurse, physician, lab_director, admin_director_finance, transfusion nurse…), Drizzle/PostgreSQL, Eden Treaty (type-safe API), next-intl (ru/en/uz), shadcn / glassmorphism UI.
• AI ("Cortex"): AI vositalari oilasi (patient-health-book va boshqalar uchun cortex tool), OCR pipeline, PDF korpusidan IAC ekstraksiyasi.
• Infra/deploy: GCP (Cloud Build — cloudbuild.yaml 19 KB, cloudbuild-dev.yaml, .gcloudignore), Docker, Redis, infra/.
• AI tooling (ko'p): .claude/ (243 sessiya!), .codex/, .cursor/, .opencode/, .agents/, .deep-research/, .harness/, .mcp.json, .playwright-cli — multi-agent ishlab chiqish.

Juda katta hajm; PR-driven (raqamlar #14xx). Ko'rinadigan klinik/biznes modullari:

• RCM (Revenue Cycle Management): claims, coding, denials — tab-render tuzatishlari, field-mapping, seed fixtures (#1453, #1452).
• RBAC/auth: persona-demo bo'shliqlarini yopish, kirish grantlari (rcm→admin_director_finance, checkups→lab_director+physician) (#1451, #1450, #1447).
• Checkups: UX parity, healthcare-crm'dan 5 ta surface'ni porting qilish (#1448).
• Patient Health Book: cortex tool + admin sahifasi (#1436), sidebar + queue OCR dep (#1446).
• IAC akkreditatsiya: "ASIC Composer 2.5 evidence slice", workplan control plane, ekstraksiya uchun real PDF korpus gate'i (#1449, #1444, #1442).
• Foydalanuvchi hissasi: ~760 kommit (aniq modullar uchun alohida git log --author tahlili kerak; hajm bo'yicha — asosiy jamoa a'zolaridan biri).

(Platforma miqyosida; mualliflik taqsimlangan — foydalanuvchi yagona emas, balki asosiy hissa qo'shuvchilardan biri.)

• "Bitta mahsulot = RBAC + konfiguratsiya orqali turli tenantlar" — arxitektura invarianti: parallel kod bazalari o'rniga mijozlar rollar/konfiguratsiya/Cortex routing orqali ajratiladi. → Platforma darajasidagi mahsulot-arxitektura fikrlashi.
• Chuqur klinik RBAC — medication orders (CPOE: administer-medication), transfuziya monitoringi, nazorat qilinadigan moddalar chiqindisining guvohlik ostida yo'q qilinishi — klinik vorkflowlar uchun nozik ruxsatlarga ega rollar. → Domenli murakkab avtorizatsiya (sog'liqni saqlash, regulyativ talablar).
• AI pipelinelar (Cortex/OCR/IAC): "gate" tekshiruvlari bilan real PDF korpuslaridan ekstraksiya, OCR queue, har bir bemor uchun AI vositalari. → Production ML/AI integratsiyalari.
• Multi-tenancy + GCP deploy (Cloud Build, dev/prod), type-safe monorepo (Eden Treaty), E2E botlar bilan nihoyatda yuqori PR tezligi.

• Topilgan sessiyalar: ushbu loyiha uchun lokal Claude Code sessiyalar katalogida 243 ta .jsonl — portfolioning rekordi, qolganlaridan bir necha barobar ko'p. Bundan tashqari parallel vositalar: Codex (.codex, git'dagi Codex bot), Cursor, opencode, .agents, .deep-research, .harness.
• AI bilan nima qilingan: aslida barcha ishlab chiqish AI-native; 3.5 oyda 4556 kommit faqat kuchli AI yordami bilan mumkin. CLAUDE.md (35 KB) + AGENTS.md (35 KB) — agentlar uchun batafsil strategik ko'rsatmalar.
• AI workflow patternlari (brend uchun eng yuqori daraja): multi-agent (Claude Code + Codex + Cursor), E2E botlar bilan PR-driven (US healthtech platform E2E), CLAUDE.md'dagi strategik "invariantlar", mahsulotning AI qatlami sifatida Cortex, tadqiqot va verifikatsiya oqimlari uchun .deep-research/.harness.

• 4556 kommit / ~3.5 oy; foydalanuvchi ~760 (2-muallif).
• Sog'liqni saqlash platformasi: LIS/CIS/CRM/RCM/akkreditatsiya/bemor + AI (Cortex/OCR).
• 5 ta app (web/admin/api/queue/sites), GCP deploy, chuqur RBAC.
• 243 CC sessiyasi + multi-agent tooling — portfolioning miqyosda AI-driven ishlab chiqishining eng kuchli isboti.