Nava ChatBot

• 

  sample output screen for Amazon Nova chatBot

Inspiration I wanted to see how fast AI could actually be. Most chatbots feel slow, so I built Nova ChatBot to test the low-latency capabilities of the new Amazon Nova 2 Lite model. My goal was to create a "zero-lag" assistant that feels like a natural conversation.

What it does Nova ChatBot is a responsive AI assistant that provides instant answers to any query. It features a clean chat interface and a built-in memory system, allowing it to remember previous parts of the conversation for a seamless user experience.

How we built it The app is built with Python and Streamlit for a lightweight frontend. It connects to Amazon Bedrock using the Boto3 library. I specifically used the Nova 2 Lite model via the Converse API to ensure the dialogue is fast and efficient.

Challenges we ran into The main challenge was configuring AWS permissions (IAM) to allow the app to talk to the model. I also had to troubleshoot a validation error, which I solved by switching to a US-based inference profile—a key step for accessing the newest Nova models on-demand.

Accomplishments that we're proud of I am proud of creating a fully functional AI app that is secure. By using Streamlit Secrets, I ensured that my AWS keys remain private and protected, even when the code is shared on GitHub.

What we learned I learned how to navigate the Amazon Bedrock ecosystem, specifically how to manage chat history through the API and how to optimize model performance using regional inference profiles.

What's next for Nova ChatBot I plan to add RAG (Retrieval-Augmented Generation) so the bot can answer questions based on uploaded PDF documents. I also want to explore adding multimodal features, allowing Nova to "see" and describe images.

Built With

• amazon-web-services
• amazonbedrock
• amazonnova2lite
• boto3
• github
• json
• python
• streamlit