# CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

## 🚀 Project Overview

**ROA2WEB** - Modern ERP Application with two main modules:
1. **Reports App** (`reports-app/`) - Read-only reports from Oracle (raportări)
2. **Data Entry App** (`data-entry-app/`) - Data input with approval workflow (introduceri date)

**Main Branch**: `main` (use for PRs)
**Working Directory**: Repository root

**Quick Reference**: See `README.md` for complete setup, commands, deployment, and testing instructions.

---

## 📁 Application-Specific Instructions

> **IMPORTANT**: When working on a specific application, ALWAYS read its dedicated CLAUDE.md first!

| Application | CLAUDE.md Location | Description |
|-------------|-------------------|-------------|
| **Data Entry** | `data-entry-app/CLAUDE.md` | Bonuri fiscale, chitanțe, workflow aprobare |
| **Reports** | This file (below) | Rapoarte Oracle read-only |
| **Telegram Bot** | `reports-app/telegram-bot/README.md` | Bot Telegram |

### When to Use Which Instructions

**Working on `data-entry-app/`**:
→ **FIRST read `data-entry-app/CLAUDE.md`** for:
  - SQLModel + Alembic patterns (NOT Oracle)
  - SQLite database (NOT Oracle pool)
  - Workflow states (DRAFT → PENDING → APPROVED)
  - Receipt/Attachment/AccountingEntry models
  - Expense types and auto-generation logic

**Working on `reports-app/` or `shared/`**:
→ Use instructions from this file (below)

**Working on shared components** (`shared/auth/`, `shared/database/`, `shared/frontend/`):
→ These are used by BOTH apps - be careful with changes!
→ `shared/frontend/` contains: LoginView.vue, auth store factory, login styles

---

## 🏗️ Architecture

### Microservices Structure
```
.
├── shared/                    # Shared components (DB pool, auth, frontend)
│   ├── database/             # Oracle pool (used by both apps)
│   ├── auth/                 # JWT auth (used by both apps)
│   └── frontend/             # Shared Vue components, stores, styles
│       ├── components/       # LoginView.vue
│       ├── stores/           # auth.js (Pinia store factory)
│       └── styles/           # login.css
│
├── reports-app/              # READ-ONLY reports from Oracle
│   ├── backend/              # FastAPI API (port 8001)
│   ├── frontend/             # Vue.js 3 UI (port 3000-3005)
│   └── telegram-bot/         # Telegram bot (port 8002 internal)
│
├── data-entry-app/           # DATA INPUT with approval workflow
│   ├── backend/              # FastAPI API (port 8003) - SQLite + SQLModel
│   ├── frontend/             # Vue.js 3 UI (port 3010)
│   └── CLAUDE.md             # ⚠️ READ THIS for data-entry work!
│
├── docs/                     # Architecture & style guides
├── deployment/               # Production deployment scripts
└── ssh-tunnel/              # SSH tunnel for Oracle DB access
```

### Starting Services

**Quick Start** (All services with parallel backend startup):
```bash
./start-dev.sh               # Dev: Backend :8001, :8003, Bot :8002, Frontend :3000 (~11s)
./start-test.sh              # Test: Same ports (~33s - Oracle pool init takes longer)
```

**Individual Service Control** (for quick development iterations):
```bash
./frontend.sh restart        # Restart frontend only (~7s - fastest!)
./backend-reports.sh start   # Start Reports backend :8001
./backend-data-entry.sh stop # Stop Data Entry backend :8003
./bot.sh status              # Check Telegram bot :8002 status
./status.sh                  # Show all services status + health checks
```

**Infrastructure**:
```bash
./ssh_tunnel.sh start        # Oracle DB tunnel (production: 10.0.20.36)
./ssh-tunnel-test.sh start   # Oracle TEST tunnel (LXC: 10.0.20.121)
```

**Benefits**:
- **87% faster frontend restart**: 7s vs 53s full restart
- **38% faster full startup**: 33s vs 53s (test) via parallel backend init
- **Granular control**: Restart individual services without affecting others

### Key Architectural Decisions
- **Shared Database Pool**: Singleton `OraclePool` in `shared/database/oracle_pool.py` (python-oracledb with connection pooling)
- **Centralized Auth**: JWT-based auth in `shared/auth/` with middleware auto-injecting `request.state.user`
- **Two-Tier Cache System**: Hybrid L1 (Memory) + L2 (SQLite) cache in `backend/app/cache/` - **MANDATORY for all new endpoints**
- **SSH Tunnel**: Required for Oracle DB connections (development/Linux) - see Database Setup
- **FastAPI Structure**: Services in `backend/app/services/` with `@cached` decorator, routers in `backend/app/routers/`, models in `backend/app/models/` or `schemas/`
- **Telegram Bot**: Standalone SQLite database for bot data, communicates with backend via HTTP API

---

## 🗄️ Database Setup

**Schema**: `CONTAFIN_ORACLE` (authentication and user management)
**Connection**: SSH tunnel required (Oracle on remote network)

### SSH Tunnel (Development/Linux)
```bash
./ssh_tunnel.sh start     # localhost:1526 → remote:1521
./ssh_tunnel.sh status    # Check tunnel
./ssh_tunnel.sh stop      # Stop tunnel
```

**IMPORTANT**: Always ensure SSH tunnel is running before starting backend services.

### Environment Variables (`reports-app/backend/.env`)
```bash
# Oracle Database (through SSH tunnel)
ORACLE_USER=CONTAFIN_ORACLE
ORACLE_PASSWORD=your_password
ORACLE_HOST=localhost
ORACLE_PORT=1526
ORACLE_SID=ROA

# JWT Authentication
JWT_SECRET_KEY=your_secret_key
JWT_ALGORITHM=HS256
JWT_EXPIRE_MINUTES=30

# Telegram Bot Integration
TELEGRAM_BOT_INTERNAL_API=http://localhost:8002
```

**Windows Production**: Direct Oracle connection, no SSH tunnel required. Ensure `TELEGRAM_BOT_INTERNAL_API` is set for auth code management.

---

## 🔑 Authentication Flow

1. **Login**: `POST /api/auth/login` → calls `pack_drepturi.verificautilizator(username, password)`
2. **Token**: JWT includes `username`, `user_id`, `companies[]`, `permissions[]`, `exp`, `iat`, `type`
3. **Middleware**: `AuthenticationMiddleware` in `shared/auth/middleware.py` validates tokens, injects user
4. **Protected Routes**: All routes except `excluded_paths` require valid JWT

**Key Files**:
- `shared/auth/middleware.py` - FastAPI middleware with rate limiting (5 req/5 min)
- `shared/auth/jwt_handler.py` - Token creation/validation
- `reports-app/backend/app/main.py` - Auth router inline definition

---

## 📝 Common Development Tasks

### Adding a New API Endpoint
**IMPORTANT**: Always use the cache system for database queries to improve performance.

1. Create **service** in `reports-app/backend/app/services/your_service.py` (NOT in router!)
2. Define Pydantic schemas in `app/schemas/` or `app/models/`
3. **Add caching** using `@cached` decorator in service methods
4. Create router in `reports-app/backend/app/routers/your_router.py` (calls service)
5. Register router in `app/main.py`: `app.include_router(your_router, prefix="/api/your_prefix")`

**Service Example with Caching** (RECOMMENDED):
```python
# app/services/your_service.py
from app.cache.decorators import cached
from database.oracle_pool import oracle_pool

class YourService:
    @staticmethod
    @cached(cache_type='schema', key_params=['company_id'])
    async def _get_schema(company_id: int) -> str:
        """Get schema for company (CACHED 24h)"""
        async with oracle_pool.get_connection() as connection:
            with connection.cursor() as cursor:
                cursor.execute("""
                    SELECT schema FROM CONTAFIN_ORACLE.v_nom_firme
                    WHERE id_firma = :company_id
                """, {'company_id': company_id})
                result = cursor.fetchone()
                return result[0] if result else None

    @staticmethod
    @cached(cache_type='your_data', key_params=['filter_params', 'username'])
    async def get_your_data(filter_params: YourFilter, username: str) -> YourResponse:
        """
        Get your data from Oracle (CACHED 10 min)

        Cache automatically:
        - Generates unique key from filter_params + username
        - Stores in L1 (memory) + L2 (SQLite)
        - Returns cached data on subsequent calls
        - Tracks performance metrics
        """
        schema = await YourService._get_schema(filter_params.company_id)

        async with oracle_pool.get_connection() as connection:
            with connection.cursor() as cursor:
                cursor.execute(f"""
                    SELECT * FROM {schema}.your_table
                    WHERE your_condition = :param
                """, {'param': filter_params.param})
                rows = cursor.fetchall()
                # Process results...
                return YourResponse(data=processed_data)
```

**Router Example** (calls service):
```python
# app/routers/your_router.py
from app.services.your_service import YourService

@router.get("/", response_model=YourResponse)
async def get_your_data(
    filter_params: YourFilter = Depends(),
    current_user: CurrentUser = Depends(get_current_user)
):
    """Get your data - delegated to service with caching"""
    return await YourService.get_your_data(filter_params, current_user.username)
```

**Cache Configuration** (add to `app/cache/config.py` if new cache type):
```python
# Add TTL for your cache type
ttl_your_data: int = int(os.getenv('CACHE_TTL_YOUR_DATA', '600'))  # 10 min default

# Add to get_ttl_for_type() method:
'your_data': self.ttl_your_data,
```

**Cache Best Practices**:
- ✅ Use `@cached` decorator for ALL database queries
- ✅ Place logic in services (NOT routers)
- ✅ Cache schema lookups separately (long TTL: 24h)
- ✅ Choose appropriate TTL (frequently changing data: 5-10 min, static data: 30 min - 24h)
- ✅ Include `username` in `key_params` for user-specific data
- ✅ Include filter parameters in `key_params` for query variations
- ❌ Don't query Oracle directly in routers (use services with caching)
- ❌ Don't skip caching for performance-critical endpoints

### Adding a New Frontend Page/Component
**IMPORTANT**: Follow the established CSS architecture and design system.

**Before writing ANY CSS**: Read **`docs/ONBOARDING_CSS.md`** (5-minute quick start) → See "Documentation Index" below for complete guide list.

**Golden Rules**:
- ✅ Use global patterns first (`.roa-card`, `.roa-metric`, `.roa-badge-*`) - check `CSS_PATTERNS.md`
- ✅ Use design tokens (`var(--color-primary)`) not hardcoded values (`#2563eb`)
- ✅ **Use shared CSS** from `src/assets/css/` - NEVER create new CSS when shared classes exist
- ✅ For inline stats/totals use `.summary-stats-inline`, `.stat-item`, `.stat-label`, `.stat-value` from `stats.css`
- ❌ Never use `:deep()` in components (use `src/assets/css/vendor/` for PrimeVue overrides)
- ❌ Never duplicate CSS (write once, use everywhere)
- ❌ Never add new scoped CSS for patterns that already exist in shared CSS files

**Tables - Unified Column Structure & Filter Buttons**:
- ✅ **ALWAYS use separate columns** for related data (Debit | Credit, not Debit+Credit stacked)
- ✅ **Use PrimeVue DataTable** with one value per `<Column>` component
- ✅ **Add filter/action buttons** (clear, export Excel, export PDF, refresh) **on separate row below filters**
- ✅ **PrimeVue Button** components with **icon + label** (not icon-only!)
- ✅ **Export ALL data** from backend (page_size: 999999), not just current page
- ❌ **Never group multiple values** vertically in a single column
- ❌ **Never use HTML `<button>` for filter actions** (use PrimeVue `<Button>`)
- ❌ **Never put action buttons on same row as filters** (separate row!)
- ❌ **Never export only current page** (must fetch all data for export)
- 📖 **See**: `CSS_PATTERNS.md` → Table Patterns → Unified Table Column Structure & Filter/Action Buttons

### Adding a New Telegram Bot Command
**IMPORTANT**: Follow established command patterns and formatting.

**Before coding**: Read **`reports-app/telegram-bot/TELEGRAM_COMMANDS.md`** for command patterns → See "Documentation Index" for complete guides.

**Standard Pattern** (add handler in `app/bot/handlers.py`):
```python
async def your_command(update: Update, context: ContextTypes.DEFAULT_TYPE) -> None:
    telegram_id = update.effective_user.id

    # 1. Check authentication (use helpers.py)
    user = await get_user_by_telegram_id(telegram_id)
    if not user:
        await update.message.reply_text("⚠️ Account not linked. Use /start with code.")
        return

    # 2. Check active company (use session.py)
    active_company = await get_active_company(telegram_id)
    if not active_company:
        await update.message.reply_text("⚠️ Select company first: /selectcompany")
        return

    # 3. Call backend API (use api/client.py)
    # 4. Format response (use formatters.py)
    # 5. Add tests + update MANUAL_TESTING_CHECKLIST.md
```

### Adding Shared Functionality
1. Place in `shared/{database|auth|utils}/`
2. Import using `sys.path.append()` pattern (see `backend/app/main.py:21`)
3. Add tests in `shared/tests/`

### Frontend API Integration
```javascript
// Store pattern (src/stores/yourStore.js)
import axios from 'axios';

const api = axios.create({
  baseURL: 'http://localhost:8001/api',
  headers: { 'Authorization': `Bearer ${token}` }
});

const response = await api.get('/endpoint');
```

---

## 🔒 Security

- **Git Hooks**: Security scanning in `security/install_hooks.sh`
- **Environment Files**: Never commit `.env` (use `.env.example` as template)
- **SSH Keys**: In `ssh-tunnel/secrets/` (gitignored)
- **JWT Secrets**: Generate strong secrets for production
- **Rate Limiting**: Built into auth middleware (5 requests per 5 minutes)

---

## 📚 Documentation Index

### Quick Start & Commands
→ **`README.md`** - Project overview, setup, development commands, testing, deployment

### Data Entry App (Bonuri Fiscale)
- **`data-entry-app/CLAUDE.md`** - ⚠️ **READ FIRST** when working on data-entry
- `data-entry-app/README.md` - Quick start guide
- `docs/data-entry/REQUIREMENTS.md` - Functional requirements
- `docs/data-entry/ARCHITECTURE.md` - Technical architecture (SQLModel, workflow)

### Architecture & Planning
- **`docs/ARCHITECTURE_SCHEMA.md`** - Architecture diagrams, cache system, and schemas
- `docs/MICROSERVICES_GUIDE.md` - Microservices architecture details
- `DEVELOPMENT_BLUEPRINT.md` - Detailed development plan

### Frontend Development
- **`docs/ONBOARDING_CSS.md`** - CSS system quick start (START HERE for new components)
- **`docs/CSS_PATTERNS.md`** - Complete CSS patterns library (cards, forms, buttons, etc.)
- **`docs/DESIGN_TOKENS.md`** - Design tokens (colors, spacing, typography)
- `docs/STYLING_GUIDELINES.md` - CSS best practices and conventions
- `docs/COMPONENT_STYLING.md` - Component-specific styling guide
- `docs/FORM_TEMPLATE.md` - Standardized form template
- `reports-app/frontend/README.md` - Frontend architecture and setup
- `reports-app/frontend/tests/README.md` - Playwright E2E testing guide

### Backend Development
- `reports-app/backend/README.md` - Backend architecture and API details
- API endpoints documented in `README.md` (Authentication, Companies, Dashboard, Invoices, Treasury, Telegram)

### Telegram Bot Development
- **`reports-app/telegram-bot/README.md`** - Complete bot architecture and development guide (START HERE)
- **`reports-app/telegram-bot/TELEGRAM_COMMANDS.md`** - Command reference and patterns
- `tests/MANUAL_TESTING_CHECKLIST.md` - Manual testing procedures

### Deployment
- **`DEPLOYMENT_GUIDE.md`** - Production deployment (Linux/Docker & Windows/IIS)
- `deployment/windows/README.md` - Windows deployment quick start
- `deployment/windows/docs/WINDOWS_DEPLOYMENT.md` - Complete Windows guide
- `deployment/windows/docs/TELEGRAM_BOT_TROUBLESHOOTING.md` - Bot troubleshooting

### Troubleshooting
→ **`README.md`** - Common issues (SSH tunnel, backend, frontend, database)
→ **Backend**: Check `.env`, SSH tunnel status, port 8001 availability
→ **Frontend**: Clear node_modules, check Node.js version (16+), Vite auto-ports
→ **Database**: Verify SSH tunnel, Oracle listener, credentials, test `/health`

---

## 🔧 Tech Stack

**Backend**: FastAPI, python-oracledb, JWT (PyJWT), Pydantic, pytest
**Frontend**: Vue.js 3 (Composition API), PrimeVue, Pinia, Vite, Axios, Chart.js, Playwright
**Telegram Bot**: python-telegram-bot, SQLite + aiosqlite, httpx, FastAPI (internal), pytest
**Infrastructure**: Oracle Database, SSH Tunnel, Nginx (Linux prod), Docker (Linux prod), IIS + NSSM (Windows prod)

---

**For detailed information on any topic, always check the Documentation Index above first.**