Inspiration
Looking at replays to improve gameplay is often a grueling and time consuming task, so we built an end of year fun coop game to improve.
What it does
The app generates scenarios based on weaknesses gleaned from player data by AI agents.
Challenges we ran into
Real-time State Synchronization: Managing the shared state of a multiplayer lobby over WebSockets was a significant challenge. Ensuring that player actions (like readiness or making a choice), disconnections, and server-side events (like the scenario timer expiring) were all handled gracefully and reflected instantly for all users required careful state management and locking mechanisms on the backend.
Agent Reliability and Prompt Engineering: Getting the AI agent to reliably use its tools and provide structured, accurate feedback was an iterative process. We invested significant time in crafting detailed system prompts that guide the agent to query our databases and interact with our map MCP server.
Accomplishments that we're proud of
We are proud of everything we created in this project!
Built With
- alb
- amazon-web-services
- amplify
- beanstalk
- bedrock-agent-core
- bedrock-knowledge-base
- cloudwatch
- ecs
- fastapi
- lambda
- rds
- redshift-serverless
- riotapi
- s3
- sagemaker

Log in or sign up for Devpost to join the conversation.