DiscountMate

Updated version: Trimester 1, 2026
Organisation: DataBytes
Repository: DataBytes-Organisation/DiscountMate_new

DiscountMate is a capstone grocery price intelligence platform that helps Australian shoppers compare supermarket prices, track savings, manage grocery lists, discover specials, and access AI-supported shopping assistance. The project combines full-stack application development, data engineering, machine learning, data analytics, cyber security, and cloud deployment work into one integrated product.

The T1 2026 version focuses on making DiscountMate more production-ready through Cloud Run deployment, CI/CD automation, ML sidecar services, a custom domain, Profile Hub improvements, and stronger support for live showcase demonstrations.

Live Application

Service	URL
Custom domain	https://discountmate.app/
Frontend Cloud Run	https://webdev-frontend-518391291595.australia-southeast1.run.app/
Backend Cloud Run	https://webdev-backend-518391291595.australia-southeast1.run.app/
API docs	https://webdev-backend-518391291595.australia-southeast1.run.app/api-docs

Current live deployment flow:

discountmate.app
-> Firebase Hosting
-> Cloud Run frontend
-> Cloud Run backend
-> ML sidecars inside backend service

Project Goals

DiscountMate aims to:

• Help users compare grocery prices across major Australian retailers.
• Surface weekly specials and discount opportunities.
• Support smarter shopping lists and basket comparison.
• Provide personalised user hub features such as profile management, notifications, subscriptions, support, and privacy pages.
• Use ML/AI features for recommendations, OCR receipt extraction, reverse image search, price prediction, and recipe assistance.
• Provide a deployable cloud architecture that can be maintained by future project teams.

Key Features

• Product browsing, searching, and filtering.
• Category-based product discovery.
• Product detail and price comparison views.
• Basket and grocery list management.
• Smart list and list reprice support.
• Weekly specials powered through backend and ML endpoints.
• Reverse image search for product lookup.
• OCR receipt extraction support.
• Recipe assistant / RAG chatbot.
• Profile Hub with dashboard, notifications, alert segments, profile management, subscription, support, and privacy/terms pages.
• Support request submission to the project support email.
• Custom domain for showcase-ready access.

Repository Structure

DiscountMate_new/
|-- Backend/                    # Node.js / Express backend API
|-- Backend/ml-service/          # Python Flask ML service
|-- Backend/ml-service/reverse_image_search/
|   `--                         # FastAPI reverse image search sidecar
|-- Frontend/                   # Expo / React Native Web frontend
|-- DE/                         # Data Engineering ETL pipeline work
|-- ML/                         # ML research, notebooks, and experiments
|-- Data/                       # Data resources and local data references
|-- docs-site/                  # Docusaurus documentation site
|-- discount-mate-infra/        # Infrastructure as code / cloud setup references
|-- Catalogue_Scraping_2025/    # Catalogue OCR and scraping support
|-- Documentation/              # Project documentation
|-- .github/workflows/          # GitHub Actions workflows
|-- firebase.json               # Firebase Hosting rewrite to Cloud Run frontend
`-- README.md                   # Project overview and setup guide

Technology Stack

Area	Technology
Frontend	Expo, React Native Web, TypeScript
Backend	Node.js, Express, MongoDB
ML service	Python, Flask, Gunicorn
Reverse image search	Python, FastAPI, Uvicorn, FAISS
OCR	Tesseract OCR and Python image processing
Recipe assistant	RAG pipeline, OpenRouter-compatible LLM integration
Data Engineering	Python, DuckDB, PostgreSQL, ETL pipelines
Deployment	Docker, Google Cloud Run, Artifact Registry
Hosting	Firebase Hosting custom domain rewrite
CI/CD	GitHub Actions
Documentation	Docusaurus, Markdown

Cloud Architecture

The current Cloud Run architecture keeps the frontend public and the backend as the single API entrypoint.

User browser
-> discountmate.app
-> Firebase Hosting
-> webdev-frontend Cloud Run service
-> webdev-backend Cloud Run service
   -> backend container, port 8080
   -> ml-service sidecar, port 5001
   -> reverse-image-search sidecar, port 8001

The ML containers are not exposed publicly. The backend calls them through localhost inside the same Cloud Run service:

ML_SERVICE_URL=http://127.0.0.1:5001
REVERSE_IMAGE_SEARCH_SERVICE_URL=http://127.0.0.1:8001

This keeps the public API surface smaller while still allowing the frontend to access ML-backed features through normal backend routes.

Main Services

Frontend

Location:

Frontend/

The frontend is an Expo / React Native Web app. It is built as a web export and deployed to Cloud Run. The deployed frontend reads the backend API URL from:

EXPO_PUBLIC_API_URL

Example:

EXPO_PUBLIC_API_URL=https://webdev-backend-518391291595.australia-southeast1.run.app/api

Backend

Location:

Backend/

The backend is an Express API that handles product, user, basket, list, notification, support, ML proxy, reverse image search proxy, and Swagger documentation routes.

Important deployed backend routes:

GET  /
GET  /api-docs
GET  /api/products
GET  /api/categories
GET  /api/ml/weekly-specials
POST /api/ml/recommendations
POST /api/ml/price-prediction
POST /api/contact
GET  /api/reverse-image-search/health

ML Service

Location:

Backend/ml-service/

The ML service provides:

• Weekly specials endpoint.
• Recommendation endpoint.
• Price prediction endpoint.
• OCR receipt processing.
• Recipe RAG endpoints.

It runs on port 5001 in deployment.

Reverse Image Search

Location:

Backend/ml-service/reverse_image_search/

Reverse image search runs as a dedicated FastAPI sidecar on port 8001. It loads FAISS assets from Google Cloud Storage in Cloud Run.

Local Development Setup

Prerequisites

Install:

• Node.js 20+
• npm
• Python 3.10 or 3.11
• Docker Desktop
• Google Cloud CLI, only required for deployment tasks
• Firebase CLI, only required for Firebase hosting/domain tasks

Clone the repository:

git clone https://github.com/DataBytes-Organisation/DiscountMate_new.git
cd DiscountMate_new

If working from a fork, add the organisation repository as upstream:

git remote add upstream https://github.com/DataBytes-Organisation/DiscountMate_new.git
git fetch upstream

Backend Local Setup

cd Backend
npm ci
cp .env.example .env
npm start

Default local backend:

http://localhost:3000

Backend scripts:

npm start
npm run lint
npm run check:ci

Required local environment variables are documented in:

Backend/.env.example

Minimum local backend values:

PORT=3000
BASE_URL=http://localhost:3000
MONGO_URI=<mongodb connection string>
JWT_SECRET=<local jwt secret>
CORS_ORIGIN=http://localhost:8081
ML_SERVICE_URL=http://localhost:5001
REVERSE_IMAGE_SEARCH_SERVICE_URL=http://localhost:8001

Support email variables:

SUPPORT_EMAIL=supportdiscountmate@gmail.com
SUPPORT_EMAIL_USER=supportdiscountmate@gmail.com
SUPPORT_EMAIL_APP_PASSWORD=<gmail app password>

Do not commit .env files or real secrets.

Frontend Local Setup

cd Frontend
npm ci
cp .env.example .env
npm run web

Default local frontend:

http://localhost:8081

Frontend scripts:

npm run web
npm run lint
npm run test:ci
npm run build:web

Frontend environment:

EXPO_PUBLIC_API_URL=http://localhost:3000/api

For live builds:

EXPO_PUBLIC_API_URL=https://webdev-backend-518391291595.australia-southeast1.run.app/api

ML Service Local Setup

cd Backend/ml-service
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python app.py

Default local ML service:

http://localhost:5001

Useful checks:

curl http://localhost:5001/health
curl "http://localhost:5001/api/weekly-specials?limit=2"

Some ML features require additional environment variables:

GOOGLE_CLOUD_PROJECT=sit-26t1-discountmate-935cb94
FAISS_BUCKET_NAME=discountmate-ml-models
FAISS_OBJECT_NAME=reverse_image_search.faiss
RAG_BUCKET_NAME=discountmate-ml-models
RAG_OBJECT_PREFIX=recipe_rag/
OPEN_ROUTER_API_KEY=<secret>
HUGGING_FACE_TOKEN=<secret>

Reverse Image Search Local Setup

Reverse image search can be run locally as a separate FastAPI service:

cd Backend/ml-service
source venv/bin/activate
uvicorn reverse_image_search.api:app --host 0.0.0.0 --port 8001

Default local reverse image search service:

http://localhost:8001

Health check:

curl http://localhost:8001/health

If using GCS assets locally, authenticate with Google Cloud:

gcloud auth application-default login
gcloud config set project sit-26t1-discountmate-935cb94
gcloud auth application-default set-quota-project sit-26t1-discountmate-935cb94

Running the Full App Locally

Use three terminals.

Terminal 1, ML service:

cd Backend/ml-service
source venv/bin/activate
python app.py

Terminal 2, backend:

cd Backend
npm start

Terminal 3, frontend:

cd Frontend
npm run web

Then open:

http://localhost:8081

CI/CD Workflow

Primary workflow:

.github/workflows/app-dev-cloud-run-prep.yml

The workflow has two main purposes:

• Validate PRs with frontend, backend, and ML checks.
• Deploy frontend and backend services to Cloud Run after changes are merged to main, when deployment is enabled.

Deployment is gated by:

ENABLE_CLOUD_RUN_DEPLOY=true

The workflow builds and pushes images for:

• webdev-frontend
• webdev-backend
• webdev-ml-service
• webdev-reverse-image-search

Images are pushed to Artifact Registry:

australia-southeast1-docker.pkg.dev/sit-26t1-discountmate-935cb94/discount-mate-images

Required GitHub Actions Variables

Set in:

GitHub repository -> Settings -> Secrets and variables -> Actions -> Variables

Required variables:

ENABLE_CLOUD_RUN_DEPLOY=true
GCP_PROJECT_ID=sit-26t1-discountmate-935cb94
GCP_REGION=australia-southeast1
GCP_ARTIFACT_REGISTRY_REPO=discount-mate-images
BACKEND_CLOUD_RUN_SERVICE=webdev-backend
FRONTEND_CLOUD_RUN_SERVICE=webdev-frontend
BACKEND_IMAGE_NAME=webdev-backend
FRONTEND_IMAGE_NAME=webdev-frontend
ML_IMAGE_NAME=webdev-ml-service
REVERSE_IMAGE_SEARCH_IMAGE_NAME=webdev-reverse-image-search
BACKEND_PUBLIC_API_URL=https://webdev-backend-518391291595.australia-southeast1.run.app/api
CORS_ORIGIN=https://webdev-frontend-518391291595.australia-southeast1.run.app,https://discountmate.app,https://www.discountmate.app
FAISS_BUCKET_NAME=discountmate-ml-models
FAISS_OBJECT_NAME=reverse_image_search.faiss
RAG_BUCKET_NAME=discountmate-ml-models
RAG_OBJECT_PREFIX=recipe_rag/

Optional variables:

BASE_URL=https://webdev-backend-518391291595.australia-southeast1.run.app
ANALYTICS_SERVICE_URL=<analytics service url if available>

Required GitHub Actions Secrets

Set in:

GitHub repository -> Settings -> Secrets and variables -> Actions -> Secrets

Required secrets:

GCP_SA_KEY
MONGO_URI
JWT_SECRET
OPEN_ROUTER_API_KEY
HUGGING_FACE_TOKEN
SUPPORT_EMAIL
SUPPORT_EMAIL_USER
SUPPORT_EMAIL_APP_PASSWORD

Important:

• GCP_SA_KEY authenticates GitHub Actions to Google Cloud.
• SUPPORT_EMAIL_APP_PASSWORD must be a Gmail app password, not a normal Gmail password.
• Do not store real secrets in source files, screenshots, PR descriptions, or issue comments.

Google Cloud Deployment Notes

Main project:

sit-26t1-discountmate-935cb94

Region:

australia-southeast1

Runtime service account:

github-actions-prod@sit-26t1-discountmate-935cb94.iam.gserviceaccount.com

Current public Cloud Run services:

webdev-backend
webdev-frontend

The backend service uses multiple containers in a single Cloud Run revision:

backend                 port 8080
ml-service              port 5001
reverse-image-search    port 8001

Firebase Hosting and Custom Domain

Firebase Hosting is used to provide a clean public domain for the frontend:

https://discountmate.app/

The Firebase config rewrites all routes to the Cloud Run frontend:

{
  "hosting": {
    "public": "public",
    "rewrites": [
      {
        "source": "**",
        "run": {
          "serviceId": "webdev-frontend",
          "region": "australia-southeast1"
        }
      }
    ]
  }
}

Because the browser origin becomes https://discountmate.app, the backend CORS configuration must include:

https://discountmate.app
https://www.discountmate.app

Manual Cloud Run Validation

Useful deployed backend checks:

curl https://webdev-backend-518391291595.australia-southeast1.run.app/
curl https://webdev-backend-518391291595.australia-southeast1.run.app/api/reverse-image-search/health
curl "https://webdev-backend-518391291595.australia-southeast1.run.app/api/ml/weekly-specials?limit=2"

Check CORS from the custom domain:

curl -i -H 'Origin: https://discountmate.app' \
  'https://webdev-backend-518391291595.australia-southeast1.run.app/api/products?page=1'

Expected:

HTTP/2 200
access-control-allow-origin: https://discountmate.app

Test support email endpoint:

curl -i -H 'Origin: https://discountmate.app' \
  -F 'name=Deployment Test' \
  -F 'email=supportdiscountmate@gmail.com' \
  -F 'topic=Other' \
  -F 'subject=Cloud Run support email test' \
  -F 'message=Testing support email delivery from Cloud Run deployment.' \
  'https://webdev-backend-518391291595.australia-southeast1.run.app/api/contact'

Expected:

emailStatus: "sent"

Data Engineering

The DE/ folder contains ETL pipeline work for transforming retailer data into structured outputs used by the application and analytics work.

Current focus areas include:

• Retailer product ingestion.
• Bronze to silver transformations.
• Product identity and matching.
• Category and retailer normalization.
• Price and promotion history.
• Reusable ETL framework patterns.

See:

DE/etl-pipeline/

Documentation Site

The docs-site/ folder contains the Docusaurus documentation site.

Run locally:

cd docs-site
npm ci
npm start

Build:

npm run build

Branching and Contribution Guide

Recommended workflow:

git fetch upstream
git checkout main
git pull upstream main
git checkout -b feature/your-feature-name

Before opening a PR:

git status
git diff

Run relevant checks:

cd Backend && npm run check:ci
cd Frontend && npm run lint
cd Frontend && npm run test:ci
cd Frontend && npm run build:web

PR guidance:

• Keep PRs focused and reviewable.
• Do not commit .env files, credentials, model artefacts, or local virtual environments.
• Include screenshots or test evidence for UI and deployment changes.
• Mention any required GitHub Secrets, GCP permissions, or manual setup steps.
• For deployment changes, explain whether they were tested locally, manually on Cloud Run, or through GitHub Actions.

Security and Secret Handling

Never commit:

• .env files
• Gmail app passwords
• API keys
• MongoDB credentials
• Google Cloud service account JSON
• Hugging Face tokens
• OpenRouter keys
• large private model files

Use GitHub Secrets or Google Secret Manager for sensitive values.

If a secret is accidentally exposed:

1. Revoke or rotate it immediately.
2. Update GitHub Secrets or Secret Manager.
3. Tell the team lead or mentor.
4. Avoid reusing the exposed value.

Known Deployment Considerations

• Cloud Run reserves the PORT variable for the ingress container. Do not manually set PORT in the Cloud Run manifest.
• Docker images for Cloud Run must support linux/amd64.
• ML containers may need higher memory and longer startup probes than the backend.
• Reverse image search requires access to FAISS files in Google Cloud Storage.
• Recipe RAG and LLM features require external API keys and model/token configuration.
• Firebase custom domain traffic requires matching backend CORS origins.
• Support email requires a Gmail app password.

Troubleshooting

Frontend loads but products fail

Check:

• EXPO_PUBLIC_API_URL
• Backend Cloud Run health
• Backend CORS origin
• Browser console network errors

Backend cannot call ML service

Check:

• ML_SERVICE_URL
• ML sidecar startup logs
• Cloud Run multi-container manifest
• ML container memory and startup probe

Reverse image search fails

Check:

• REVERSE_IMAGE_SEARCH_SERVICE_URL
• FAISS GCS bucket and object names
• Runtime service account GCS permissions
• Reverse image search container logs

Support request does not email

Check:

• SUPPORT_EMAIL
• SUPPORT_EMAIL_USER
• SUPPORT_EMAIL_APP_PASSWORD
• Gmail app password validity
• Backend /api/contact response emailStatus

GitHub Actions deploy does not run

Check:

• Branch is main
• ENABLE_CLOUD_RUN_DEPLOY=true
• Required variables are set
• Required secrets are set
• GitHub Actions permissions and GCP service account permissions

Project Handover Notes

For future teams:

• Keep the custom domain and Cloud Run URLs documented.
• Keep GitHub Actions variables and secrets up to date.
• Rotate exposed or outdated secrets.
• Continue using small PRs for deployment changes.
• Validate production-like changes manually before relying on automation when external services are involved.
• Keep ML assets out of Git unless they are intentionally small and approved for version control.
• Add evidence to PRs when changing deployment, security, or user-facing behaviour.

Contact and Support

Project support email:

supportdiscountmate@gmail.com

For repository access, deployment permissions, or GCP issues, contact the current DataBytes project leads, mentors, or assigned technical administrators.

Acknowledgement

DiscountMate is developed as a Deakin University DataBytes capstone project. The T1 2026 version reflects contributions across App Development, Data Engineering, Machine Learning, Data Analytics, Cyber Security, Infrastructure, and Documentation teams.