Planning Overview - huqianghui/AI-Coach-vibe-coding GitHub Wiki
Project Overview
Auto-generated from
.planning/PROJECT.md
Last synced: 2026-04-02
AI Coach Platform — BeiGene
What This Is
A production AI coaching platform for BeiGene (百济神州) that trains Medical Representatives (MRs) through AI-simulated HCP interactions. MRs practice F2F calls and conference presentations with digital Healthcare Professionals, receive multi-dimensional scoring, and track their improvement over time. Built on Azure PaaS services with i18n support for global deployment (China + Europe).
Core Value
MRs can practice realistic conversations with AI-powered digital HCPs and receive immediate, multi-dimensional feedback to improve their communication skills and product knowledge — anytime, without needing a real HCP or trainer.
Requirements
Validated
- ✓ Project skeleton with FastAPI backend + React frontend — existing
- ✓ AI adapter subsystem with pluggable provider pattern (Azure OpenAI, Claude, Mock) — existing
- ✓ Database layer with async SQLAlchemy + Alembic migrations — existing
- ✓ Docker Compose deployment configuration — existing
- ✓ CI/CD pipeline with GitHub Actions — existing
- ✓ JWT authentication with User model, RBAC (admin/user/manager), login/me endpoints — Phase 1
- ✓ Design system with Figma Make tokens and 17 shadcn/ui components — Phase 1
- ✓ Pluggable AI service adapters (STT/TTS/Avatar) with mock implementations — Phase 1
- ✓ React SPA with i18n (zh-CN/en-US), auth store, router with guards — Phase 1
- ✓ Responsive layouts (user top-nav, admin sidebar), login page — Phase 1
- ✓ Feature toggle config API and frontend ConfigProvider — Phase 1
Active
- F2F HCP coaching with chat and voice interaction
- Conference presentation mode with virtual HCP audience
- Multi-dimensional scoring and feedback system
- Training session lifecycle management
- HCP profile configuration (personality, knowledge, interaction rules)
- Scenario management (products, key messages, scoring weights)
- Training material management (upload, versioning, retention)
- Personal and organizational reports/dashboards
- Azure OpenAI integration (GPT-4o + Realtime model)
- Azure Speech Services (STT/TTS)
- Azure AI Avatar (digital human for HCP)
- Azure Content Understanding (multimodal evaluation)
- Azure service configuration UI
Out of Scope
- Teams Bot integration — deferred to post-MVP, architecture should allow it
- OAuth / Azure AD SSO — future, use simple auth for now
- WeChat Mini Program — future, responsive web covers mobile for now
- Multi-tenancy — single tenant per-region deployment
- Real-time video conferencing — simulated conference, not live video
- Mobile native app — responsive web-first
Context
- Client: BeiGene (百济神州) — major biotech company, needs global deployment
- Reference: Adapted from Capgemini AI Coach for AWS solution (see
docs/capgemini-ai-coach-solution.md) - Architecture patterns: Reuse from two reference projects:
- ragflow-skill-orchestrator-studio (Connection management, agent adapters)
- yoga-guru-copilot-platform (ServiceConfig dual-layer, UI components, multi-provider agents)
- Existing codebase: Skeleton exists but most modules are empty stubs. Starting fresh implementation, keeping the project structure.
- Design: Figma Design System created (Figma Make). Individual page prompts in
docs/figma-prompts/. - UI reference: Capgemini screenshots in
pdf/images/(mobile-first, adapting to web) - Timeline: Prototype needed this week (week of 2026-03-24) for client demo
Constraints
- Cloud: Azure PaaS only (no AWS) — Azure OpenAI, Speech, Avatar, Content Understanding, PostgreSQL
- i18n: Must support Chinese + English from day 1, i18n framework required for European expansion
- Compliance: Per-region deployment to satisfy data residency regulations (China, EU)
- Auth: Simple user/admin for MVP, architecture must support Azure AD (Entra ID) later
- Budget: Azure AI Avatar is premium — implement as configurable option, fall back to Azure Speech TTS
- Frontend: Must be responsive — same app works on desktop, tablet, mobile, and Teams Tab
Key Decisions
| Decision | Rationale | Outcome |
|---|---|---|
| Azure PaaS over AWS | Client requirement, BeiGene uses Azure | — Pending |
| Start fresh, reuse patterns | Existing code is skeleton stubs, cleaner to rebuild with proven patterns | — Pending |
| Figma-first design | User designs in Figma, code generated from Figma MCP | — Pending |
| Simple auth for MVP | Speed to demo, Azure AD integration later | — Pending |
| GPT Realtime + Speech fallback | Premium voice experience with cost fallback option | — Pending |
| i18n from day 1 | European expansion planned, retrofitting i18n is costly | — Pending |
Evolution
This document evolves at phase transitions and milestone boundaries.
After each phase transition (via /gsd:transition):
- Requirements invalidated? → Move to Out of Scope with reason
- Requirements validated? → Move to Validated with phase reference
- New requirements emerged? → Add to Active
- Decisions to log? → Add to Key Decisions
- "What This Is" still accurate? → Update if drifted
After each milestone (via /gsd:complete-milestone):
- Full review of all sections
- Core Value check — still the right priority?
- Audit Out of Scope — reasons still valid?
- Update Context with current state
Last updated: 2026-03-31 after Phase 11 completion — HCP Profile Agent Integration (auto-create AI Foundry agents on HCP CRUD, agent sync status badges, table UI, token broker per-HCP agent_id)