# Visual AI QC - AI-Powered Quality Control Platform

AI-driven visual quality control for marketing materials and design assets using OpenAI GPT-4o and Google Gemini 2.5 Pro. Analyzes images and video against brand guidelines through **75 QC checks**, **14 profiles**, serving **8 clients**.

## Work in `backend/` Directory

**All development and production code is in the `backend/` directory.**

## Directory Structure

```
ai_qc/
+-- backend/                  # MAIN APPLICATION
|   +-- api_server.py         # Flask application
|   +-- llm_config.py         # LLM configuration (GPT-4o, Gemini 2.5 Pro)
|   +-- ocr_measurement.py    # Tesseract OCR layout measurements
|   +-- profile_config.py     # Profile management
|   +-- client_config.py      # Client configuration (8 clients)
|   +-- pdf_processor.py      # PDF brand guideline processing
|   +-- media_plan_processor.py # Excel media plan parsing
|   +-- usage_tracker.py      # Usage tracking and cost estimation
|   +-- visual_qc_apps/       # QC check modules (75 checks)
|   +-- profiles/             # QC profile configurations (14 profiles)
|   +-- brand_guidelines/     # Reference assets and guidelines DB
|   +-- scripts/              # Deployment and testing scripts
|   +-- config/               # Environment configs
|   +-- uploads/              # File uploads (production)
|   +-- output/               # Generated reports (production)
|   +-- uploads-dev/          # File uploads (development)
|   +-- output-dev/           # Generated reports (development)
|   +-- usage_logs/           # Usage tracking JSONL logs
|   +-- media_plans/          # Uploaded media plans
|   +-- web_ui.html           # Web interface
|   +-- README.md             # Full documentation
|   +-- CLAUDE.md             # Development guide
|
+-- frontend/                 # Frontend-only deployment (optional)
+-- CLAUDE.md                 # Top-level development guide
+-- README.md                 # This file
```

## Quick Start

### Local Development
```bash
cd backend/
./scripts/run-local.sh
# Access at http://localhost:7183
```

### Deploying to Dev / Prod

Code changes ship through Bitbucket branches (`develop` → dev server, tagged releases on `main` → prod server). Deploys are run manually on each server:

```bash
ssh <server>
cd /opt/ai_qc
backend/scripts/deploy.sh dev              # dev — pulls latest develop
backend/scripts/deploy.sh prod v1.2.0      # prod — checks out a specific tag
backend/scripts/deploy.sh dev --dry-run    # preview without applying
backend/scripts/rollback.sh last           # revert the most recent deploy
backend/scripts/health-check.sh            # quick "is it alive?"
```

See `backend/CLAUDE.md` for the full environment / branch / deploy reference.

## Environments

| Env | URL | Branch | Status |
|---|---|---|---|
| Local | `http://localhost:7183` | any | For active development |
| Dev | `https://optical-dev.oliver.solutions/ai_qc/` | `develop` | Live |
| Prod | `https://optical-prod.oliver.solutions/ai_qc/` | `main` (tagged) | Not yet stood up |

## Clients

Diageo, Unilever, L'Oreal, Amazon, Boots, Dow Jones, Honda, General

## Capabilities

- **Image QC**: 75 checks across brand compliance, typography, layout, accessibility
- **Video QC**: Native Gemini video analysis (visual quality, brand consistency, text legibility, pacing)
- **OCR Measurements**: Tesseract pixel-level data for margin and alignment checks
- **PDF Guidelines**: Multi-page PDF upload with automatic LLM summarization
- **Media Plans**: Excel media plan parsing with asset spec validation
- **Enterprise Auth**: Azure AD with MSAL/PKCE and silent token refresh
- **User Access Control**: Default-deny per-user client access with admin grant/revoke and audit log
- **Admin Panel**: Usage analytics, user tracking, cost estimation, access management
- **Profile Versioning**: Automatic version control on profile edits

## Documentation

- **Full Documentation**: See `backend/README.md`
- **Development Guide**: See `backend/CLAUDE.md`
- **Deployment Guide**: See `backend/DEV_PROD_SETUP.md`