• 

  Main Screen

• 

• 

  Final generated image

• 

• 

  Rating sceen

• 

  Launchdarkly feature flags

Convex + TypeScript + Bedrock

Inspiration

We are climbers who enjoy putting our problem-solving skills into physical activity. Everyone wants to learn something new, but it can initially be intimidating. That is why we have Climbing Route Setter, to help the setters an optimal and multi-step according to the desired difficulty.

What it does

It helps to generate climbing routes based on the difficulty and type selected. Users can then rate the router generated once they have climbed them to further train the AI on what constitutes a good route

Challenges we ran into

Learning new tools like Launch Darkly, Convex, and AWS Bedrock.

Using the model to generate accurate images for our task. We ran into a few issues with bedrock/convex integration. The main blocker was a recursive json bug that prevented the bedrock API from running inside the convex backend - this was fixed by a very helpful convex engineer though :)

Accomplishments that we're proud of

We used all the technologies available in the hackhathon:

• WE used LaunchDarkly to manage the possible options for climbing routes
• Bedrock (Titan) to generate climbing routes
• Convex to store images and their ratings and interact with Bedrock API

Working application with Convex to run with.

What we learned

Time management :)

Launchdarkly flags are cool actually AWS Bedrock

Convex

Images into base64 string

What's next for Climbing Route Setter

Having someone to try it out and set the climbing pieces together! Use more launchdarkly flags to switch between Bedrock Models

Built With

• bedrock
• convex
• launch-darkly
• react
• ts
• vite