Los Valientes

LLM-powered digital assistant with Amazon Bedrock to build teams and answer questions about VALORANT Esports players.

Comment

About the Project

Inspiration

The inspiration for this project stemmed from my passion for esports and the dynamic, fast-paced world of Valorant. The Valorant VCT hackathon provided the perfect opportunity to combine my interest in competitive gaming with cutting-edge technology, specifically leveraging Amazon Bedrock to create an intelligent chat assistant.

What it does

Our digital assistant helps in scouting and recruitment for Valorant esports teams. It provides team compositions based on various prompts, assigns player roles, and justifies team effectiveness. The assistant leverages real-time data from various sources to make informed recommendations.

How we built it

  1. Data Collection: Aggregated textual data from APIs, Valorant Wiki, and S3 buckets.
  2. Development: Utilized streamlit library to build the chat interface, incorporating amazon Bedrock functionalities to answer questions about Valorant esports players and team compositions.
  3. Testing: Conducted try and error testing to ensure accurate responses, refining the information retrieval process.
  4. Deployment: Deployed the assistant into a user-friendly interface for real-time interaction on a AWS ECR.

Challenges we ran into

  • Finding Enough Data: Data provided by the S3 buckets lacked alot of important features, especially agent related features. We had to look for other data sources to retrive player specific performance per agent.
  • Data Integration: Merging data from multiple sources while ensuring accuracy and consistency.
  • Textual Data: Parsing the xml files was no easy task, we had to experiment in a juypter notebook to extract features we deemed important.

Accomplishments that we're proud of

  • Successfully built an AI-powered assistant capable of making strategic team recommendations.
  • Developed a user-friendly interface that enhances the scouting and recruitment process for Valorant esports teams.

What we learned

  • How to deal with 100 gbs of json files.
  • Effective use of Amazon Bedrock for building intelligent applications.
  • AWS services and how to use them.

What's next for Los Valientes

We plan to:

  • Expand the assistant's capabilities to cover more aspects of esports management.
  • Add more textual data to feed the hungry LLM.
  • Extract more features that can be used in evaluating the models performance.
Share this project:

Updates