[Syed.Atyab@Codeline.rihal.om] One-Click Secure Image Caption Demo

[CodelineAtyab]

Omantel Hackathon – Secure Image Captioning Demo

Goal:
Enable participants to log in with their corporate Outlook account, upload a single image, and instantly receive an AI-generated caption—demonstrating the complete secure stack:

Azure OAuth2 ➜ FastAPI ➜ n8n ➜ HuggingFace

---

User Story

As a…	I want…	So that…
Omantel hackathon participant	to log in with my corporate Outlook account and upload one image to receive an AI-generated caption	I can validate the end-to-end secure stack in a single, easy demo

---

Acceptance Criteria

1. SSO & Authentication
   • Unauthenticated visits to / are redirected to Microsoft Azure SSO.
   • After successful login, a valid JWT is stored in the FastAPI session.
   • FastAPI verifies the JWT signature using Azure’s public key.
   • No password fields are ever displayed in the UI.

2. Main Page UI
   • Displays a simple form:

   • Image-file selector (max 5 MB)
   • “Generate Caption” button

3. Image Upload & Storage
   • On button press, the image is uploaded to MinIO/S3.
   • File URL + user email are inserted into Postgres.

4. Caption Generation Workflow
   • FastAPI triggers the Image-to-Caption n8n workflow via internal API (secret API-key header).
   • n8n calls a HuggingFace model to generate a text caption.
   • Caption is saved back to the same Postgres row.
   • Caption is immediately returned to the frontend and displayed beneath the uploaded image.

5. Performance
   • Full round-trip (upload → caption → display) completes in ≤ 10 seconds for images ≤ 5 MB on a developer laptop via docker-compose.

6. Persistence & History
   • Refreshing the page lists the user’s last 3 captioned images, retrieved from Postgres.

7. Observability
   • Prometheus metrics exposed by FastAPI at /metrics:

   • image_upload_total
   • caption_latency_seconds
     • Metrics are viewable in a pre-configured Grafana dashboard.

8. Local Development
   • All services (FastAPI, n8n, Postgres, MinIO) run locally via the provided docker-compose.yml.
   • A single command starts everything:

   docker compose up

---

Tech Stack Overview

Layer	Technology	Purpose
Authentication	Azure OAuth2	Corporate SSO & JWT issuance
API / Backend	FastAPI	Session handling, API gateway, metrics
Workflow Orchestration	n8n	Image-to-Caption pipeline
AI Model	HuggingFace	Caption generation
Object Storage	MinIO/S3	Image storage
Database	Postgres	Metadata & captions
Monitoring	Prometheus + Grafana	Metrics & dashboards
Containerisation	Docker Compose	One-command local setup

---

Quick Start

# Clone repository
git clone <repo-url> && cd <repo-dir>

# Launch the entire stack
docker compose up

Open your browser at http://localhost:8000, sign in with your Outlook account, and start captioning!