AWSpeaks: Talk to Your Cloud

Inspiration

A lot of times, simple questions like "Does this bucket exist in my account?" or "If yes, where is it located?" require logging into AWS, grabbing an MFA device, and navigating the console—tedious for quick answers. I wanted a way to ask these questions naturally, directly via Slack or chat, and get instant responses without all the hassle.

What it does

Users can ask natural language questions about their AWS resources, such as how many Lambda functions they have, or if a specific S3 bucket exists and where. The assistant instantly provides clear answers without needing to log into the AWS console.

How I built it

I used Amazon Lex to build the chatbot and handle intent understanding, AWS Lambda for processing queries and intelligently routing requests to the appropriate AWS SDK calls, and Amazon Bedrock’s large language models (Claude v2) for natural language processing and query classification.

Challenges I ran into

Maintaining session context across multi-turn conversations was tricky to ensure smooth interactions. Crafting the right prompts for Bedrock to get precise responses for dispatching service handlers was another challenge that required careful tuning.

What I learned

Bedrock’s language models are incredibly powerful. By combining them with AWS Lambda’s serverless compute, I built capabilities that would have seemed impossible a decade ago—turning natural language queries into actionable AWS API calls seamlessly.

What's next for AWSpeaks: Talk to Your Cloud

I plan to expand support for more AWS services and add a broader set of operations, making it an even more comprehensive conversational assistant for cloud management.

Built With

• amazon-web-services
• bedrock
• cdk
• lambda
• lex
• python