This project provides a multilingual, AWS-native generative AI platform for building chatbots. It targets developers and organizations looking to leverage Amazon Bedrock for conversational AI, RAG, bot sharing, and task automation, offering a scalable and managed solution.

How It Works

The platform is built on AWS managed services, including Amazon Bedrock for foundational models, Amazon OpenSearch Serverless for RAG vector search, and Amazon Cognito for user authentication. It supports chat, custom bots with RAG capabilities, a bot store for sharing, and agent-based task automation. The architecture emphasizes serverless components like API Gateway, Lambda, CloudFront, and S3 for a scalable and infrastructure-light deployment.

Quick Start & Requirements

• Install/Run: Clone the repository and execute ./bin.sh in AWS CloudShell.
• Prerequisites: AWS account, configured AWS CLI, and model access enabled in us-east-1 for Bedrock.
• Setup Time: Approximately 35 minutes for deployment.
• Docs: Migration Guide, Local Development

Highlighted Details

• Supports Retrieval-Augmented Generation (RAG) with importable Knowledge Bases.
• Features a bot store for sharing custom bots and API publishing for standalone bots.
• Includes agent functionality for complex task automation and tool integration.
• Offers extensive configuration options via cdk.json or parameter.ts for security, multi-environment deployments, and feature toggles.

Maintenance & Community

The project is maintained by AWS Samples. Contribution guidelines are provided, encouraging feature request issues before pull requests for enhancements.

Licensing & Compatibility

Licensed under the MIT-0 License, permitting commercial use and modification without significant restrictions.

Limitations & Caveats

Version 3 requires a migration from V2, with V2 bots becoming unusable without it. For production, disabling self-registration and restricting IP ranges is strongly recommended due to security implications. Lambda SnapStart availability may vary by region for Python functions.