Build multimodal search with Amazon OpenSearch Service

This article provides a step-by-step guide for building a multimodal search solution using Amazon OpenSearch Service and the Amazon Bedrock Titan Multimodal Embeddings model. It enables search capabilities for both text and images, and combinations of the two.

Specifically, the article covers:

• Multimodal search solution architecture
• Prerequisites and data overview
• Step 1: Creating the OpenSearch-Amazon Bedrock ML connector
• Step 2: Creating the OpenSearch ingest pipeline with the text_image_embedding processor
• Step 3: Creating the k-NN index and ingesting the retail dataset
• Step 4: Performing multimodal search experiments (text-only, image-only, text+image)
• Observations and conclusions from the experiments