💡 Built using Google Cloud AI + Database integration to demonstrate real-world Generative AI applications.
NeighborLoop is an AI-powered web application that enables intelligent surplus sharing within a community. It integrates Gemini 3 Flash with AlloyDB AI to analyze items, generate natural language descriptions, and enable semantic search directly within the database.
This project was implemented as part of the Google Cloud Codelab: Gemini 3 Flash + AlloyDB Sustainability App and deployed using Google Cloud services.
👉 https://neighbor-loop-5808901341.us-central1.run.app
NeighborLoop is designed as a swipe-style surplus-sharing platform where users can explore items shared within a community.
Instead of relying on manual descriptions or keyword-based search, the system uses AI to:
- Understand item images and text
- Automatically generate meaningful descriptions
- Enable semantic, context-aware discovery
-
🧠 AI-Generated Descriptions Gemini analyzes images and produces human-like descriptions
-
🔍 Semantic Search Finds relevant items based on meaning rather than exact keywords
-
🗄️ AlloyDB AI Integration Embeddings and vector operations executed directly inside the database
-
📸 Multimodal Processing Combines image and text understanding
-
☁️ Cloud-Native Deployment Built and deployed using Google Cloud (Cloud Shell + managed Cloud Run)
-
💡 Interactive UI Swipe-inspired interface for exploring items
Frontend
- HTML, CSS (Templates)
Backend
- Python (Flask)
Database
- AlloyDB (PostgreSQL with vector extensions)
AI
- Gemini 3 Flash API
- Text Embeddings (
text-embedding-005)
Cloud
- Google Cloud Platform (GCP)
- Cloud Shell (Development Environment)
- Cloud Run (Managed Deployment)
- Cloud Storage (Image Handling)
-
Items (with images) are uploaded or displayed
-
Images are stored in Cloud Storage
-
Gemini API:
- Analyzes images
- Generates descriptive text
-
AlloyDB:
- Stores structured data
- Generates embeddings within SQL
-
Semantic queries retrieve similar items based on meaning
-
Results are presented through a simple UI
- Gemini 3 Flash → Multimodal AI reasoning
- AlloyDB AI → Vector storage + semantic querying
- Cloud Storage → Image storage
- Flask Backend → Application logic
- Cloud Run → Managed deployment service
├── templates/ # Frontend templates
├── app.py # Flask application
├── requirements.txt # Dependencies
├── README.md # Documentation
└── .gitignore
This application was deployed using Google Cloud Shell and integrated Google Cloud services.
- Enabled required GCP APIs
- Created and configured AlloyDB instance
- Set up Cloud Storage bucket
- Configured environment variables in Cloud Shell
- Ran and tested the application in cloud environment
- Deployed and accessed the application via Cloud Run service URL
- Configured Google Cloud environment and enabled required APIs
- Set up and connected AlloyDB instance
- Integrated Gemini 3 Flash API for AI-generated descriptions
- Worked with and configured a Flask-based backend application
- Configured Cloud Storage for image handling
- Deployed and tested the application using Cloud Shell and Cloud Run
- Integrating Generative AI with databases
- Using Gemini for multimodal AI applications
- Implementing semantic search using embeddings
- Working with AlloyDB AI capabilities
- Deploying applications using Cloud Run and Cloud Shell
This project demonstrates how modern applications can combine:
- Generative AI (Gemini)
- Vector databases (AlloyDB AI)
- Cloud-native deployment (GCP)
to build intelligent, scalable, and real-world solutions.
This project is based on a Google Cloud Codelab and demonstrates real-world integration of AI with cloud-native databases and services.
- User authentication system
- Personalized recommendations
- Improved UI/UX
- Mobile responsiveness
Apache 2.0 License
Varshini Gurram AI | Cloud | Data Enthusiast Building intelligent and scalable applications
If you found this project useful, consider giving it a ⭐ on GitHub!