20250307 AWS Generative AI Hackathon Warm-up Workshop

demo for workshop on 250307 @awseducatestdambtw

Key Concepts

• Multi-Modal Foundation Model
• Opensearch serverless

Introduction

This project integrates multimodal model technology and utilizes the generative AI capabilities of Amazon Bedrock to create a web application. Users can obtain recommended images through the following methods:

1. Text Search: Enter keywords to retrieve relevant images.
2. Image Search: Upload an image to find similar ones.

The system retrieves image data from OpenSearch based on user input and provides personalized recommendations, creating an image exploration experience similar to Pinterest.

Key Technologies and Services

1. Amazon Bedrock

• Uses amazon.titan-embed-image-v1 for embedding images and text to enhance search accuracy.

2. OpenSearch

• Creates an index to store and retrieve image tags and descriptions.
• Provides high-performance full-text and vector-based search.

3. Amazon S3

• Stores static frontend pages and image data.
• Supplies image URLs for OpenSearch indexing.

4. AWS Lambda & API Gateway

• Connects the AI chatbot backend.
• Provides API services for frontend interaction with OpenSearch.

Features Overview

• Keyword Search: Retrieves images based on tags and descriptions via OpenSearch.
• Image Similarity Search: Uses Amazon Bedrock embedding models to match similar images.
• Personalized Recommendations: Provides personalized image suggestions based on user search history and interactions.

Deployment and Usage

Prerequisites

• AWS services must be enabled (Bedrock, OpenSearch, S3, Lambda, API Gateway).
• Sufficient IAM permissions to manage the mentioned AWS services.

Deployment Steps

1. Set Up OpenSearch Index

   • Configure image tags and description fields.
   • Upload initial image data.

2. Configure Amazon S3

   • Upload images and static frontend files.
   • Allow OpenSearch to access image URLs.

3. Set Up Amazon Bedrock

   • Enable amazon.titan-embed-image-v1 model.
   • Create an API endpoint for Lambda integration.

4. Deploy Lambda and API Gateway

   • Create a Lambda function to handle search requests.
   • Expose API via API Gateway for frontend access.

references

Image gallery

• pixabay
• kaggle

Inspiring Websites

• https://www.seopital.co/blog/pinterest-keyword-tool