Inspiration
To gain insight in the latest Technologies of AI and the feel of working in a many pieces environment to integrate an AI assistant using RAG/LLM.
What it does
** Valorant Q Bot**
As the name suggests, a general AI assistant for a esports managers to select it pro players for the team
Using next js frontend web application ,typescript ,python,json/jquery, to integrate with amazon bedrock foundation models,lambda,amazon opensearch,and lang chain agents with implementation required multiple configuration across multiple services and applying policies and to Provide a gen -ai assistant for valorant game to select the pro players in a esports league/tournament.
Using game decided,total scores,configuration.
How we built it
Frontend-Client-nextjs/vercel sdk/AWS SDK Backend Server AWS Bedrock Models Claude -3-Sonnet
Challenges we ran into
segregating data for each of the leagues
Challenge was to analyze the huge volume of data across many leagues conducted for a series of three years and being scalable and to accommodate relevant data in the knowledge base, pre-processing and filter irrelevant metadata to avoid latency and ease in data retrieval and response and applying the logic for the required inference.
Removing observers target, metadata from the each of the games files using jq filters. uploading data consumed most of the development time, -tried on local machine-encountered SSL certificate error, -tried from cloudshell -but not sufficient storage/execution scenario, -tried batch operation -failed many times due to insufficient memory, -finally uploaded with glue. Keeping a watch on the cost of services and ensuring to keep it that it doesn't cross the credits .(unfortunately went above the credits )
Accomplishments that we're proud of
- Was able to create knowledge base two days before deadline
What we learned
Many under the hood things to explore from AWS environment. Both technical and pricing/budgeting/cost reduction.
What's next for Valorant Esports Manager
Dashboard Metrics
Built With
- awsbedrock
- awsglue
- awslambda
- awsopensearchserverlessservice
- awss3
- base
- jq
- knowledge
- node.js
- python
- typescript
Log in or sign up for Devpost to join the conversation.