This document explains how to run the backend API and frontend UI for this project.

• Windows PowerShell
• Python virtual environment at .venv
• Node.js and npm installed

Open a new terminal at project root.

cd C:\Users\JHASHANK\Downloads\VERISIGHT_V1
.\.venv\Scripts\Activate.ps1
python -m uvicorn engine.pipeline.app:app --host 127.0.0.1 --port 8000 --reload

Backend URLs:

• API base: http://127.0.0.1:8000
• API docs: http://127.0.0.1:8000/docs
• Verify endpoint used by frontend: http://127.0.0.1:8000/api/v1/verify

Open another terminal.

cd C:\Users\JHASHANK\Downloads\VERISIGHT_V1\frontend
npm install
npm run dev

Open the Vite URL shown in terminal (usually http://localhost:5173).

• Input: product image only
• The UI auto-sends today date for order_date and delivery_date
• mfg_date_claimed is not sent, so OCR can infer manufacturing date from the image

Cause: backend is not running.

Fix:

1. Start backend using the command in section 1.
2. Keep backend terminal open while using frontend.

Install API dependencies inside .venv:

python -m pip install uvicorn fastapi python-multipart

If you run npm commands from root accidentally, move to frontend folder first:

cd C:\Users\JHASHANK\Downloads\VERISIGHT_V1\frontend

Run a benchmark report with metrics + latency + calibration:

python evaluation/evaluate_system.py --output-json evaluation/latest_benchmark.json --append-history

Key outputs:

• evaluation/latest_benchmark.json
• evaluation/benchmark_history.json

Check regression status from saved benchmark history:

python evaluation/check_regression.py --history-json evaluation/benchmark_history.json

Calibrate decision thresholds from a benchmark report that includes per-sample data:

python evaluation/calibrate_thresholds.py --input-json evaluation/latest_benchmark.json --output-json evaluation/calibrated_thresholds.json

Environment variables:

• VERISIGHT_MAX_CONCURRENT_REQUESTS (default 4)
• VERISIGHT_REQUEST_TIMEOUT_MS (default 15000)
• VERISIGHT_OCR_ENGINE (auto, easy, paddle, yolo)
• VERISIGHT_OCR_GPU (true / false)