EcoAgent Pro: Vision-Driven Sustainability Assistant

• 

  EcoAgent Pro: Bridging the gap between environmental awareness and real-world action using Amazon Nova 2 Lite.

• 

  User Interface: A clean, Streamlit-powered dashboard hosted on AWS EC2 for global accessibility.

• 

  Multimodal Input: Processing a real-world image of produce (test.jpg) to initiate the sustainability analysis loop.

• 

  Carbon Impact Intelligence: Using Nova 2 Lite to calculate production and transportation footprints via visual reasoning.

• 

  Agentic Action: The final output providing dynamic links to local recycling facilities, moving beyond analysis into action.

• 

  Raw Data: The original test.jpg file used to demonstrate the agent's high-accuracy object recognition capabilities.

🌍 Inspiration

We wanted to simplify sustainability. Most carbon footprint tools are manual and tedious. We asked: What if you could just take a photo and get an instant action plan? EcoAgent Pro was born to turn passive observation into active environmental stewardship.

🛠️ How We Built It

The project is built on a robust AWS stack designed for scale and security:

• Amazon Bedrock & Nova 2 Lite: The "brain" of our agent, providing high-speed multimodal analysis of images and text.
• Amazon EC2: We deployed a live web server on a t3.micro instance to ensure our dashboard is accessible to anyone, anywhere.
• Streamlit: Used to create an intuitive frontend that handles image uploads and displays real-time AI reasoning.
• Boto3: The bridge between our Python logic and AWS high-performance models.

🧠 Challenges We Overcame

Our biggest hurdle was Identity and Access Management (IAM). We initially encountered a NoCredentialsError when the server tried to call Bedrock. Rather than taking the insecure route of hardcoding keys, we implemented a least-privilege IAM Role (AmazonBedrockFullAccess) and attached it directly to our EC2 instance. This ensured our agent is both functional and production-secure.

🎓 What We Learned

We learned that "Agentic AI" isn't just about answering questions—it's about providing the "next step." By integrating a direct Action Link to recycling centers based on visual recognition, we moved beyond a chatbot into a true utility.

Built With

• amazonbedrock
• amazonec2
• amazonnova
• amazonnova2lite
• awsiam
• boto3
• python
• streamlit