Bedrock-AI-News-Assistant

• 

  architecture

The idea for Bedrock AI News Assistant came from the overwhelming speed and volume of information in our daily lives. News, social media, and AI updates appear faster than we can process — so I wanted to build an intelligent assistant that could collect, summarize, and analyze only the news that truly matters to me.

The project is powered by Amazon Bedrock AgentCore, which enables a reasoning-driven agent to invoke external tools dynamically. When a user enters a query (for example, “What’s new about Tesla?”), the system automatically:

Calls a get_news_tool to fetch the latest relevant news.

Invokes an analysis agent to extract insights and summarize the key points.

To make the system more context-aware, I also implemented a custom get_time_tool, allowing the agent to fetch the current date and time. This ensures that the assistant retrieves the most recent and relevant news instead of outdated articles.

(Note: Due to environment limitations during the hackathon, the current demo runs locally, but the runtime logic and architecture mirror the live Bedrock AgentCore setup.)

Throughout the process, I learned how to integrate AWS Bedrock, AgentCore Runtime, and real-time streaming into a unified chatbot interface. One of the main challenges was handling streamed responses in Streamlit, ensuring messages appeared smoothly while maintaining session context.

This project taught me the power of tool orchestration with LLMs — bridging the gap between reasoning and real-world actions.

Built With

• bedrock
• bedrockagentcore
• python