Instantly score and improve your product catalogs using AI. Upload any PDF, get a detailed quality breakdown across five dimensions, and export a full report.
CatalogIQ reads a PDF catalog and uses an LLM to evaluate it across five independent quality dimensions. Each dimension receives a score from 0–100. The platform identifies specific issues (ranked by severity), lists concrete recommendations, and generates a downloadable PDF report — all in one workflow.
Works with HuggingFace (free), OpenAI, or a local Ollama instance.
| Home | Upload & Analyze |
 |
 |
| Results Dashboard | Analysis History |
 |
 |
| Side-by-Side Comparison | Settings |
 |
 |
| Feature | Details |
|---|---|
| 5-Dimension AI Scoring | Content Quality, Readability, Structure, Product Information, and Formatting — each 0–100 |
| Issue Detection | Severity-ranked issues (High / Medium / Low) with a targeted fix suggestion for each |
| Visual Dashboards | Circular score gauge, radar chart, and bar chart generated client-side |
| Catalog Comparison | Select any two past analyses and get a dimension-level diff with a winner |
| PDF Report Export | Download a formatted ReportLab report with scores, issues, and recommendations |
| Analysis History | Every run persisted in SQLite — browse, re-read, or delete any time |
| Multi-Provider LLM | Swap between HuggingFace, OpenAI GPT, or Ollama with a single env variable |
| Drag-and-Drop Upload | PDF files up to 50 MB, validated before processing |
| Docker Support | Spin up frontend + backend in one docker-compose up |
| REST API + Swagger | Full API at /api with interactive docs at /api/docs |
Content Quality — Grammar, factual accuracy, completeness of information
Readability — Sentence clarity, language accessibility, reading ease
Structure & Org. — Logical flow, section hierarchy, consistent categorization
Product Information — Specs, pricing, SKUs, dimensions, and descriptions
Formatting Quality — Visual consistency, layout coherence, typographic uniformity
The overall score is the equal-weighted mean of the five dimensions.
graph TD
Browser -->|HTTP / REST| API[FastAPI Backend]
API --> PDF[pdfplumber — text extraction]
API --> LLM[LangChain — HuggingFace / OpenAI / Ollama]
API --> DB[(SQLite — analysis history)]
API --> Report[ReportLab — PDF reports]
Browser --> UI[React 18 + Vite + Tailwind CSS]
UI --> Charts[Recharts — radar, bar, gauge]
| Layer | Technology | Version |
|---|---|---|
| Frontend | React, Vite, Tailwind CSS | 18 / 5 / 3.4 |
| Charts | Recharts | 2.12 |
| Routing | React Router | v6 |
| HTTP Client | Axios | 1.7 |
| File Upload | react-dropzone | 14 |
| Backend | FastAPI, Uvicorn | 0.111 / 0.29 |
| ORM | SQLAlchemy, Pydantic v2 | 2.0 / 2.7 |
| PDF Parsing | pdfplumber | 0.11 |
| AI / LLM | LangChain + providers | 0.2 |
| Reports | ReportLab | 4.1 |
| Database | SQLite | — |
| Containers | Docker, Compose | — |
- Python 3.8+
- Node.js 20+
- A free HuggingFace token or an OpenAI key or a running Ollama instance
cd backend
cp .env.example .env # then edit .env and add your token
pip install -r requirements.txt
python run.pyAPI ready at http://localhost:8000 — interactive docs at http://localhost:8000/docs
cd frontend
npm install
npm run devcp backend/.env.example backend/.env # add your token
docker-compose up --buildEdit backend/.env to choose a provider:
# HuggingFace (default — free tier available)
LLM_PROVIDER=huggingface
HUGGINGFACE_API_TOKEN=hf_xxxxxxxxxxxxxxxx
HUGGINGFACE_MODEL=mistralai/Mistral-7B-Instruct-v0.2
# OpenAI
LLM_PROVIDER=openai
OPENAI_API_KEY=sk-xxxxxxxxxxxxxxxx
OPENAI_MODEL=gpt-3.5-turbo
# Ollama (fully local, no API key needed)
LLM_PROVIDER=ollama
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=llama2| Method | Endpoint | Description |
|---|---|---|
POST |
/api/analyses/ |
Upload a PDF and run an analysis |
GET |
/api/analyses/ |
List all analyses (paginated) |
GET |
/api/analyses/{id} |
Retrieve a full analysis result |
DELETE |
/api/analyses/{id} |
Delete an analysis |
POST |
/api/analyses/compare |
Compare two analyses side-by-side |
GET |
/api/analyses/{id}/report |
Download analysis as a PDF report |
GET |
/api/health |
Backend health and version |
GET |
/api/config |
Active LLM provider and settings |
Full interactive Swagger UI at http://localhost:8000/docs.
catalog-quality-scorer/
├── backend/
│ ├── app/
│ │ ├── main.py FastAPI application and CORS setup
│ │ ├── config.py Pydantic-settings — all env variables
│ │ ├── database.py SQLAlchemy engine and session factory
│ │ ├── models/analysis.py ORM model for stored analyses
│ │ ├── schemas/analysis.py Request / response Pydantic schemas
│ │ ├── api/routes/
│ │ │ ├── analysis.py CRUD + compare + report endpoints
│ │ │ └── health.py Health and config endpoints
│ │ └── services/
│ │ ├── pdf_service.py pdfplumber text and metadata extraction
│ │ ├── analyzer.py LLM prompt, JSON parsing, score clamping
│ │ └── report_service.py ReportLab PDF report generation
│ ├── requirements.txt
│ ├── .env.example
│ ├── Dockerfile
│ └── run.py
├── frontend/
│ ├── src/
│ │ ├── pages/ Home Analyze Results History Compare Settings
│ │ ├── components/
│ │ │ ├── analysis/ ScoreCard ScoreChart AnalysisDetails IssuesList
│ │ │ ├── compare/ CompareView with diff table
│ │ │ ├── history/ HistoryTable with select-to-compare
│ │ │ ├── layout/ Navbar Layout
│ │ │ └── upload/ UploadZone (react-dropzone)
│ │ ├── services/api.js Axios API client
│ │ └── utils/helpers.js Score colours, grades, formatters
│ ├── package.json
│ ├── vite.config.js
│ ├── tailwind.config.js
│ ├── Dockerfile
│ └── nginx.conf
├── docker-compose.yml
└── README.md
Complete rewrite of the original Catalouge-Scorer Streamlit prototype (single-file, PyPDF2, Mistral-7B). Legacy code preserved on the legacy branch.