🧠 DOCUAI – Serverless Document Q&A with AWS Bedrock & Textract

DOCUAI is a serverless AI-powered document analysis system that automatically extracts text from uploaded documents using Amazon Textract, then answers user questions about the document using AWS Bedrock (Llama 3 70B Instruct model).

It’s built entirely with AWS Lambda, API Gateway, and IAM, providing a scalable, low-maintenance pipeline for intelligent document understanding.

---

⚙️ Features

• 📄 Automatic text extraction from documents (PDFs, images, etc.) using Amazon Textract
• 💬 Question answering with AWS Bedrock Llama 3 70B Instruct
• ☁️ Serverless architecture — fully managed with AWS Lambda + API Gateway
• 🔐 Secure uploads to S3 via signed requests or API Gateway
• 🧩 Simple JSON API for integration with frontends or other services
• ⚡ Asynchronous polling for reliable Textract job completion

---

🏗️ Architecture Overview

Client (Web/App)
   ↓
API Gateway  —→  AWS Lambda (Python 3.x)
                      ├── Uploads file to S3
                      ├── Runs Textract text extraction
                      ├── Calls Bedrock (Llama 3)
                      └── Returns structured answer
   ↓
Response JSON → { question, answer, message }

---

🧰 Tech Stack

Component	Service
Language	Python 3.x
AWS AI Services	Amazon Textract, AWS Bedrock
Runtime	AWS Lambda
Gateway	Amazon API Gateway
Storage	Amazon S3
Permissions	IAM Policies

---

🚀 Quick Start

1️⃣ Create Resources

• S3 bucket – to store uploaded documents.
• Lambda function – runtime: Python 3.10+
• API Gateway – with POST endpoint integrated with Lambda.
• IAM Role – attach Textract, S3, Bedrock, and CloudWatch policies.

2️⃣ Set Lambda Environment

Make sure your Lambda has:

AWS_REGION=us-west-2

3️⃣ Deploy the Lambda Code

Paste the lambda_handler and ask_question_bedrock functions into your Lambda console.

Your lambda_handler entry point will:

• Accept the payload from API Gateway,
• Upload the file to S3,
• Extract text using Textract,
• Query Bedrock Llama 3 for an answer.