Chatbot Agent with Amazon Bedrock

This project is based on the demo presented during my talk at AWS Summit Jakarta 2024 (September 5, 2024).

The Architecture

Prerequisites

1. AWS Account: Ensure you have an active AWS account.

2. AWS CLI Configuration: Set up your AWS CLI with the necessary credentials.

3. Frontend Setup: Place your frontend web application with chatbot functionality inside the ./src/web/public directory.

4. Model Access: Make sure you already request model access the model you are going to use in your Amazon Bedrock Console in AWS us-east-1 region. .

5. Environment Variables: Configure the following in the .env file:

   AWS_REGION=us-east-1
   
   # Get your Pinecone credentials from https://www.pinecone.io/
   PINECONE_CONNECTION_STRING=<pinecone connection string>
   PINECONE_API_KEY=<pinecone API key>
   
   CRAWL_TARGET_URL=<your target URL, obtained after deploying your frontend stack or using your custom domain>
   
   FROM_EMAIL=<sender email for notifications, must be configured in Amazon SES>
   EMAIL_DOMAIN=<email domain identity, also configured in Amazon SES>

Deployment Steps

1. Clone the Repository:

   git clone <repository-url>
   cd <repository-directory>

2. Install Dependencies:

   npm install && npm install --prefix src/crawler

3. Deploy the Stacks:

   cdk deploy KnowledgeBaseStack CrawlerWebStack AgentStack QueryStack

   The deployment process will start as shown below:

4. Configure Frontend:

   • Once the deployment is complete, you will receive the chatbot API endpoint (highlighted in blue).
   • Copy the endpoint URL and update your frontend configuration.

5. Deploy Frontend:

   cdk deploy FrontendStack

   • After the deployment, you will receive a CloudFront URL to access your frontend.

Running the Application

1. Prepare Knowledgebases:

   • Load documents into Amazon Bedrock from the S3 Bucket.
   • Alternatively, you can wait for EventBridge to trigger the Lambda function to crawl your web, or manually invoke the CrawlerWebStack Lambda function.

2. Interact with the Chatbot:

   • Allow Amazon Bedrock a few minutes to process your files.
   • Test the chatbot either through your frontend or via the Amazon Bedrock Management Console.