Serverless Semantic Video Search Using a Vector Database and a Multi-Modal Generative AI Embeddings Model

This AWS sample shows how to implement semantic search across images and live video by generating multimodal embeddings with Amazon Bedrock's Titan Multimodal Embeddings model and storing and querying them in Amazon OpenSearch as a vector database. It is explicitly marked as an experimental, non-production example rather than a finished product.

The project deploys through AWS Amplify, requires an AWS account with Bedrock model access enabled and appropriate IAM permissions, and includes local development prerequisites such as the AWS CLI, Python, Node.js, and the Amplify CLI. Documentation covers account prerequisites, deployment, local build and cleanup steps, usage instructions, and a walkthrough of the overall solution architecture.