AWS IDP is an AI-powered Intelligent Document Processing solution designed for unstructured data.
Transform unstructured data into actionable insights with our advanced AI-powered IDP: Analyze documents, videos, audio files, and images with unprecedented accuracy and speed.
For a visual walkthrough of the application's key features, see the Application Demo.
Multi-Modal Unstructured Data Processing
- Document Processing Content extraction, key data summarization, and layout analysis
- Video Analysis Scene detection, chaptering, and transcript generation
- Audio Analysis Speech-to-text, speaker identification
- Image Understanding Object, scene, and text recognition
AI-Powered Automation
- Bedrock Data Automation (BDA): Fast, scalable OCR + analysis
- ReAct Agent-based Workflow: Adaptive tool orchestration for any file type
- Iterative Reasoning: Multi-step refinement for accurate outputs
Hybrid Search System
- Semantic + Keyword Search: Meaning + precision combined
- Vector Indexing with OpenSearch
- Real-time Reranking for best matches
Conversational AI Interface
- MCP Server-based Chatbot: Natural language Q&A across all content
- Contextual Conversation: Multi-turn dialogue with memory
- Domain-Specific Language Support
See Backend System Architecture Overview for details.
aws-idp/
├── packages/
│ ├── frontend/ # Next.js + React user interface
│ ├── backend/ # FastAPI + MCP Tools + ReAct agent
│ ├── infra/ # AWS CDK-based infrastructure
│ │ ├── .toml # Infrastructure configuration file (e.g., dev.toml)
│ │ ├── deploy-infra.sh # Deploys core infrastructure (VPC, S3, etc.)
│ │ └── deploy-services.sh # Optional: Deploys services like ECS and ALB
│ └── results/ # Analysis results
├── docs/ # Documentation and diagrams
└── .env # Auto-generated env varsGetting started with AWS IDP involves setting up your environment, deploying the necessary cloud infrastructure, and running the application. Choose one of the following environment setup methods.
You can set up your development environment in one of two ways. The Devcontainer method is recommended for a consistent and automated experience.
Use these scripts to deploy or remove the stack quickly without local setup. They run from AWS CloudShell and execute via CodeBuild.
| Method | Description | Guide |
|---|---|---|
| Manual Local Setup | Manually install dependencies on your local machine. For advanced users or specific needs. | Manual Setup Guide / Kor |
| Devcontainer | A fully containerized environment with all dependencies pre-installed. Requires Docker and VS Code. | Devcontainer Setup Guide / Kor |
After setting up your environment using one of the guides above, proceed with the infrastructure deployment.
- Infrastructure: packages/infra/.toml
- Env vars: .env (auto-generated)
- Infrastructure: AWS CDK, Lambda, DynamoDB, S3, OpenSearch, Step Functions, Bedrock, BDA, SQS
- Backend: Python, FastAPI, MCP Tools, ReAct Agent Pattern
- Frontend: Next.js 15, React 19, TypeScript
This project is licensed under the Amazon Software License.