HackXplore 2026 -- Starter Template

Stack: Python + FastAPI + Streamlit + OpenCV + scikit-learn + Anthropic

Quick start

# 1. Copy env and fill in your keys
cp .env.example .env

# 2. Install dependencies
pip install -r requirements.txt

# 3. Start everything
python run.py

Service	URL
Frontend	http://localhost:8501
Backend	http://localhost:8000
API docs	http://localhost:8000/docs

Structure

hackxplore-template/
├── backend/
│   ├── main.py              # FastAPI app + CORS
│   └── routes/
│       ├── health.py        # GET /health
│       ├── predict.py       # POST /predict/image, /predict/timeseries
│       └── analyze.py       # POST /analyze/  (LLM)
├── model/
│   ├── vision.py            # OpenCV anomaly detection for images
│   ├── timeseries.py        # Isolation Forest for sensor data
│   └── llm.py               # Anthropic Claude wrapper + vision
├── frontend/
│   └── app.py               # Streamlit UI (3 tabs)
├── data/                    # Drop challenge data here (gitignored)
├── pitch/
│   └── deck_template.md     # Pitch deck outline
├── run.py                   # Start backend + frontend together
└── requirements.txt

Adapt for your challenge

Zeiss / image-heavy challenge:

• Extend model/vision.py with your detection logic
• Add endpoints in backend/routes/predict.py
• Use model/llm.py::ask_with_image() to describe defects in natural language

Hitachi / sensor/timeseries challenge:

• Feed CSV data into model/timeseries.py
• Add forecasting logic (e.g. simple ARIMA or rolling stats)
• Use model/llm.py::ask() to summarize anomaly reports

Submission checklist (Sunday 10:00 AM deadline)

• Pitch deck exported as .pdf in the challenge folder
• Second PDF: project title, team names, prototype link, code link
• GitHub repo set to public
• Prototype deployed and reachable (not just localhost)
• Uploaded before 10:00 AM sharp -- no changes allowed after