Inspiration
Plain text is boring and bad for shopping
What it does
Gets rich multi-modal assets that can be embedded in chat
How we built it
We have a nextJS app where users can ask for products they are interested in. We then spin up an agent network in inngest that uses the linkup mcp server to search the web for the best websites that match the intent. We store those results in inngest's network state functionality. Then we spin up a second agent passing in the website urls. That agent using a custom built MCP server to use browserbase to navigate to those websites. The custom server then parses the accessibility tree to pull out images and return them to inngest and finally pass it along to the end user so they can enjoy a multi-modal chat experience.
See attached screenshot for diagram
Challenges we ran into
Hubris and trying to do too much
Accomplishments that we're proud of
It actually works!!!
What we learned
Friendship, MCP, and agents
What's next for Ruh-Roh Chat #29
Show prices and links to purchase
Built With
- browserbase
- inngest
- linkup
Log in or sign up for Devpost to join the conversation.