Prototype fraud-detection command center for Tunisian digital payments. FastAPI + SQLAlchemy/Neon verified; Kafka, Spark, ML, and RAG remain target architecture.
---
title: Amastan Fraud Shield Guard
sdk: docker
app_port: 7860
---
# Amastan Fraud Shield Guard
A small fraud alert review API for Tunisian digital payments.
The hosted demo runs FastAPI on Hugging Face Spaces and stores data in Neon
PostgreSQL. The same code uses SQLite locally.
The project is finished and in maintenance mode.
Release
| API reference
| Deployment
| OpenAPI
| Security
## What It Does
- Stores fraud alerts.
- Gives analysts a review queue.
- Records analyst feedback.
- Exports reviewed cases.
- Tracks model metadata and training outcomes.
- Reports drift and review metrics.
- Exposes Prometheus metrics for Grafana Cloud.
- Keeps an audit trail for important changes.
Bearer tokens separate admin and analyst access.
## Proof
The first image shows the hosted API, an authenticated Swagger request, and the
same alert in Neon PostgreSQL.
The second image shows the Grafana Cloud dashboard and the metric query behind
it.
The dashboard export is stored at
`docs/grafana-dashboard.json`.
## Architecture
```mermaid
flowchart LR
Client[API client] --> API[FastAPI]
API --> Auth[Admin / analyst tokens]
Auth --> DB[(SQLite or Neon PostgreSQL)]
DB --> Review[Alerts, feedback, audit, model metadata]
API --> Metrics[Prometheus metrics]
Metrics --> Grafana[Grafana Cloud]
```
This repository contains that deployed slice only. Earlier Kafka, Spark,
Streamlit, Ollama, ChromaDB, Kubernetes, and local monitoring experiments were
removed during final cleanup.
## Main Endpoints
| Method | Path | Purpose |
|---|---|---|
| `GET` | `/health/` | Health and version |
| `POST` | `/api/v1/alerts/add/` | Store an alert |
| `GET` | `/api/v1/alerts/review-queue/` | Read the review queue |
| `POST` | `/api/v1/feedback/` | Record analyst feedback |
| `GET` | `/api/v1/stats/` | Review statistics |
| `GET` | `/api/v1/compliance/kpis/` | Compliance indicators |
| `GET` | `/api/v1/model/training-summary` | Model training history |
| `GET` | `/metrics` | Prometheus metri …