A30 Project Workspace

This repository contains the demo-ready A30 anti-fraud assistant:

• FastAPI backend with /api/health, /api/analyze, /api/intervene, /api/report, and /api/knowledge/import
• Static frontend dashboard served from /
• Lightweight rule engine, knowledge retrieval, and evaluation script

Quick Start

1. Create and activate a virtual environment.

   python -m venv .venv
   .\.venv\Scripts\Activate.ps1

2. Install dependencies.

   pip install -r requirements.txt

3. Start the app.

   uvicorn app.backend.main:app --reload

4. Open the dashboard.

   http://127.0.0.1:8000
   

Demo Flow

1. Open the dashboard and choose a role: student, adult, or elderly.
2. Paste suspicious text into the main input. Add optional audio and image hints if you want to show multimodal support.
3. Click the main analysis button to call /api/analyze.
4. Review the risk level, scam type, evidence chips, recommendation, and the generated report preview.
5. Click the intervention button to call /api/intervene.
6. For high-risk cases with a guardian contact, the backend returns guardian_notified; otherwise it falls back to a warning or tip-only action.
7. If you want to show the report endpoint directly, call /api/report with the analysis summary payload and display the returned Markdown.

Evaluation

Run the evaluation script from the repository root:

python scripts/evaluate.py

The script prints a JSON summary with total, correct, and accuracy.

Acceptance Checklist

• GET /api/health returns {"status": "ok"}
• POST /api/analyze returns risk_level, scam_type, evidence, and recommendation
• POST /api/intervene handles the high-risk guardian flow
• POST /api/report returns a Markdown safety report
• GET / serves the dashboard and /assets/* serves the frontend files
• python scripts/evaluate.py prints a JSON evaluation summary