Drbrilliant360/tayari-ai-interview

# Tayari — AI Interview Practice Platform A monorepo: a production FastAPI backend (this root) and a Next.js frontend (`frontoffice/`), wired together for local dev behind a single port. Bilingual (EN/SW), aimed at students practicing for job interviews. ## 📍 What's built **Backend** (`app/`) — the full API is implemented and tested: | Area | Endpoints | Docs | | ----------------- | ---------------------------------------------------------------------------- | ---- | | Auth & users | register / login / refresh / me, `PUT /users/me`, `/users/me/stats` | backend_thomas.md | | Sessions | full CRUD, user-scoped, `overall_score` computed on completion | backend_thomas.md | | Questions & answers | AI question generation, multipart answer upload (`202` + background pipeline), polling, re-record | ai_pipeline_elia.md | | Feedback | per-answer feedback and session-level aggregation | ai_pipeline_elia.md | | Resumes | CRUD, AI optimisation with an accept/reject diff, PDF generation | resumes_brian.md | | Analytics | `/users/me/analytics` — category breakdown, trend, delivery averages, themes | analytics_aloyce.md | **Frontend** (`frontoffice/`) — every screen of the Tayari design, wired to the API: landing, login/register, dashboard, session setup, the camera interview flow with live processing states, session history and detail, the split-view resume builder, and profile. Bilingual EN/SW throughout. See frontend_hadija.md. **Running without API keys.** With no `OPENAI_API_KEY` the AI services fall back to deterministic local generators, and with no `S3_BUCKET` set uploads go to `MEDIA_ROOT` on disk — so the whole record → analyse → feedback loop works on a laptop with nothing configured. Set the keys for real questions, transcripts and coaching. See deployment.md. **Dev tooling:** on …