# EnCoach — Adaptive AI Learning Platform

EnCoach is a smart learning environment for individual and collaborative learning,
fully integrated with AI and equipped with intelligent, professional-grade
exercises, assessments, and e-exams.

This repository hosts the full stack:

- **Backend** — Odoo 19 + ~30 `encoach_*` modules (Python 3.12, PostgreSQL 16)
- **Frontend** — React 18 + Vite + TypeScript single-page app
- **AI layer** — OpenAI + pgvector RAG, quality-gate validation, IELTS validator
- **Ops** — Docker Compose, JWT auth with refresh tokens, Prometheus-compatible
  metrics, dynamic OpenAPI 3.0 spec

> **Canonical trees:** `backend/` (all server code) and `frontend/` (all client
> code). The legacy `new_project/` directory is deprecated — see
> [`new_project/DEPRECATED.md`](new_project/DEPRECATED.md).

---

## 1. Quickstart

### Docker (recommended)

```bash
docker compose up -d
# Odoo:     http://localhost:8069
# Frontend: cd frontend && npm install && npm run dev  (http://localhost:8080)
```

Create a new Odoo database on first visit, then install the `encoach_api`
module to pull in every `encoach_*` dependency.

### Source

Prerequisites: Python 3.12, PostgreSQL 16, Node 20, npm 10.

```bash
# Backend
./setup.sh          # creates venv, installs requirements.txt
./run.sh            # starts Odoo on :8069

# Frontend
cd frontend
npm install
npm run dev         # starts Vite on :8080, proxies /api → :8069
```

See [`MANUAL-RUN.md`](MANUAL-RUN.md) for a step-by-step walkthrough and
[`CONNECT-POSTGRES.md`](CONNECT-POSTGRES.md) for database wiring.

---

## 2. Repository layout

```
odoo19/
├── backend/
│   └── custom_addons/
│       ├── encoach_api/              REST base + JWT + OpenAPI + metrics
│       ├── encoach_core/             Users, entities, roles, permissions
│       ├── encoach_taxonomy/         Subject → Domain → Topic
│       ├── encoach_ai/               OpenAI wrapper, cefr_mapper, validator
│       ├── encoach_ai_course/        AI course/exam generation pipelines
│       ├── encoach_ai_grading/       AI grading (writing/speaking/math/IT)
│       ├── encoach_ai_media/         TTS (Polly), STT (Whisper), ELAI
│       ├── encoach_vector/           pgvector store + RAG embeddings
│       ├── encoach_exam_template/    Canonical exam + student attempt models
│       ├── encoach_scoring/          Score computation, CEFR mapping
│       ├── encoach_quality_gate/     Automated content-quality checks
│       ├── encoach_ielts_validation/ IELTS-specific validators
│       ├── encoach_adaptive/         Adaptive engine, style matcher
│       ├── encoach_lms_api/          OpenEduCat bridge + LMS endpoints
│       ├── encoach_branding/         White-label config per entity
│       └── … (other encoach_* modules)
├── frontend/
│   ├── src/
│   │   ├── pages/                    Route pages (React.lazy code-split)
│   │   ├── components/               Shared UI (shadcn/ui + custom)
│   │   ├── services/                 Thin API wrappers over fetch
│   │   ├── hooks/queries/            React Query hooks + keys
│   │   ├── lib/api-client.ts         Fetch + auto JWT refresh
│   │   └── types/                    Shared DTO types
│   └── vite.config.ts                manualChunks: react, query, charts, radix…
├── docs/
│   ├── PROJECT_SUMMARY.md            Release notes & architecture history
│   ├── adr/                          Architecture Decision Records
│   └── ENCOACH_*.md                  SRS, workflows, user stories
├── docker-compose.yml
├── Dockerfile
├── requirements.txt                  Python deps (pgvector, textstat, etc.)
└── README.md                         You are here
```

---

## 3. Architecture at a glance

```
┌────────────────────┐     HTTPS/JWT      ┌────────────────────┐
│  React SPA (Vite)  │  ───────────────►  │   Odoo 19 + FastAPI-style   │
│  - React Query     │  ◄───────────────  │   controllers (encoach_*)   │
│  - next-themes     │   X-Request-Id     └──────┬─────────────┘
│  - shadcn/ui       │                           │
└────────────────────┘                           ▼
                                          ┌──────────────┐
                                          │  PostgreSQL  │
                                          │  + pgvector  │
                                          └──────┬───────┘
                                                 │
                                                 ▼
                                          ┌──────────────┐
                                          │   OpenAI,    │
                                          │   Whisper,   │
                                          │   Polly…     │
                                          └──────────────┘
```

Every request carries an `X-Request-Id` and emits structured JSON logs.
Prometheus-compatible counters are exposed at `/api/metrics`, and the live
OpenAPI 3.0 spec is at `/api/openapi.json`.

---

## 4. Key conventions

- **Canonical response envelope** — list endpoints return
  `{ items: T[], total, page, size, data: T[] }` (see `encoach_api.controllers.base.paginated_envelope`).
- **CEFR mapping** — only `encoach_ai.services.cefr_mapper` is canonical. Do
  not reintroduce local `band_to_cefr` copies.
- **JWT tokens** — short-lived access tokens (1h) + revocable refresh tokens
  (7d). Only `access` tokens are accepted as Bearer credentials; refresh tokens
  must go through `/api/auth/refresh`. See [`docs/adr/0002-jwt-refresh-token-flow.md`](docs/adr/0002-jwt-refresh-token-flow.md).
- **RAG metadata** — vector embeddings carry `course_id`, `subject_id`,
  `entity_id`, `taxonomy`, `content_hash`. Chunking kicks in above 2 000 chars.
- **Frontend pagination** — `PaginatedResponse<T>` exposes both `items` and
  `data`. Read from `items` in new code.
- **Frontend theming** — tokens live in `frontend/src/index.css` (`:root` and
  `.dark`). Always use `hsl(var(--token))` instead of raw hex.

---

## 5. Health, observability, docs

| Endpoint | Purpose |
|---|---|
| `GET /api/health` | Liveness (always 200 when server is up) |
| `GET /api/health/ready` | Readiness (DB + required config) |
| `GET /api/openapi.json` | Dynamic OpenAPI 3.0 spec generated from `@http.route` |
| `GET /api/metrics` | Prometheus-format counters per route |

---

## 6. Deployment

Staging and production both use Docker Compose. The staging server rebuilds
automatically from `main`; never force-push. See
[`INSTALL-ODOO-SUMMARY.md`](INSTALL-ODOO-SUMMARY.md) for bootstrap notes.

| Service | Staging URL |
|---|---|
| Odoo backend | `http://5.189.151.117:8069` |
| React frontend | `http://5.189.151.117:3000` |

The `.env` file is **never** committed. On staging it lives at
`/opt/encoach/backend-v2/.env`.

---

## 7. Further reading

| Document | Description |
|---|---|
| [`docs/PROJECT_SUMMARY.md`](docs/PROJECT_SUMMARY.md) | Release notes + architecture history |
| [`docs/adr/`](docs/adr/) | Architecture Decision Records (why we built it this way) |
| [`docs/ENCOACH_UNIFIED_SRS.md`](docs/ENCOACH_UNIFIED_SRS.md) | Unified frontend + backend SRS |
| [`docs/ENCOACH_ODOO19_BACKEND_SRS.md`](docs/ENCOACH_ODOO19_BACKEND_SRS.md) | Backend SRS v3.0 |
| [`docs/ENCOACH_WORKFLOWS_BACKEND_SRS.md`](docs/ENCOACH_WORKFLOWS_BACKEND_SRS.md) | Backend workflows |
| [`docs/ENCOACH_WORKFLOWS_FRONTEND_SRS.md`](docs/ENCOACH_WORKFLOWS_FRONTEND_SRS.md) | Frontend workflows |

---

## 8. Contributing

1. Branch from `main` — never push direct. Branch protection enforces it.
2. Run `npx tsc --noEmit -p tsconfig.app.json` (frontend) and the module smoke
   tests before opening a PR.
3. Every architectural decision should be captured as an ADR under
   [`docs/adr/`](docs/adr/). Copy `0000-template.md` to start one.
4. Open the PR against `main` and request review from devops (Talal).