GenAI Document Search Accelerator

Objective

This project is aimed at implementing and operationalizing an AI-powered document search solution using Azure AI Search, Azure OpenAI, and other Azure services. The goal is to introduce improvements to the GenAI document search solution and prepare it for deployment in production.

Key Objectives:

• Introduce pre-defined abstractive summaries for documents.
• Update summaries based on user search terms.
• Expand detailed summarizations after user prompts.
• Annotate based on bibliographical terms.
• Help operationalize the solution to MVP/Pilot stage in preparation for production.

---

Architecture Overview

The architecture includes the following services:

• Azure Data Lake: Stores documents and knowledge scan files.
• Azure OpenAI: Handles text summarization and embedding generation.
• Azure AI Search: Manages indexing and querying document chunks.
• Azure Cosmos DB: Stores knowledge scan metadata and references to files.
• Azure Functions: Orchestrates the process, including document chunking, generating embeddings, creating knowledge scans, and generating .docx files.

---

Deployment Instructions

Infrastructure Deployment

• Use the deploy.bicep template to deploy the required services.
• After creating the resource group, run the following command using Azure Cloud Shell:

  az deployment group create --resource-group <resource-group-name> --template-file deploy.bicep