Inspiration
Many of us make decisions when we are not fully informed or don't understand what we are buying, including all the buzzwords that appear on the page. So I created an extension that sits in the bottom right corner and is context-aware of the current page, allowing users to ask whether something is a good deal, what specific terms mean, and other relevant questions.
What it does
A context-aware AI assistant that helps you understand and evaluate financial decisions while browsing banking and shopping websites. It analyzes page content to provide relevant explanations, price comparisons, and personalized advice. A key feature is its context awareness - it analyzes the actual page content rather than relying on pretrained knowledge, which helps prevent hallucination and ensures accurate, page-specific responses.
Just ask:
- Is this a good deal here?
- What do the terms on this page mean?
- Is this reasonable APR%?
- How should I spell something
- What could be alternatives for this product?
Try it out:
- Press on this link to download it: chrome-extension.zip
- Unzip it
- Enable Developer mode in your browser
- Press Load unpacked
- Select
chat-buddydirectory - It should appear on websites like amazon.co.uk, capitalone.co.uk, houseoffraser.co.uk
How I built it
- Build a backend with Python and FastAPI
- Implemented backend logic with LLMs to make some recommendations based on user's context
- Created a bash script to bundle it up and zip it for lambda
- Configured CDK for deployment
- Bought a domain name,
- For extension I started with implementing simple chat with vanilla JS, html and css;
- Used
chrome.storage.localto store K number of previous chat messages (so they wouldn't be lost if you accidentally close a windows) - Configured SSL certificates with AWS Certificate Manager to use custom domains with API Gateway
- Created a bash script to zip it up
- Added additional endpoint in API to serve this .zip from lambda
Challenges I ran into
- CSS
- Serving .zip (At first it was sending it as ASCII instead of binary :D)
Accomplishments that I'm proud of
- Built a neat looking extension in under 24 hours
- Developed context-aware functionality that effectively interprets current window content
- Set up a complete deployment pipeline for extension distribution and AWS backend API
- Added local storage for chat history so your messages stay saved even if you close your browser
What I learned
- How to configure SSL certificates with AWS Certificate Manager and use custom domains with API Gateway
- How to serve files via AWS lambda
What's next for ChatBuddy
- Authentication!
Built With
- amazon-web-services
- bash
- cdk
- fastapi
- javascript
- lambda
- pydantic
- python
Log in or sign up for Devpost to join the conversation.