Gamerag

Inspiration

We were inspired by the curator in the book "Ready Player One". A person in the game who knows or holds information regarding the game that can help players navigate through the game and achieve objectives.

What it does

Our RAG agent is basic but has scope for tons of customization. Basically, the agent takes in context from associated "game information" files carefully crafted by game developers and answers player queries that can help guide them through the game's questline or answer general questions. This can help the player stay in-game and explore rather than jump to YouTube or the internet for a solution.

How we built it

We created a structured document for RAG (Retrieval-Augmented Generation) using Cortex Search for retrieving relevant context. This document contains carefully curated "game information" crafted by developers to help players. We integrated this with Mistral’s large language model to process queries and generate precise answers. Using effective prompt engineering, we fine-tuned the interaction to provide a seamless wizard-like experience for gamers.

Challenges we ran into

One of the main challenges was data availability—finding or creating high-quality and relevant game data for retrieval. Additionally, there was a steep learning curve in understanding the API documentation and limited resources for debugging. However, we managed to overcome these hurdles through persistence and experimentation.

Accomplishments that we're proud of

We’re proud of successfully building a functional and responsive Game Wizard within a limited time frame. The wizard outputs enhanced, contextually relevant results, providing an immersive and helpful experience for gamers. The integration of Mistral LLM and RAG worked seamlessly, which was a big win for us.

What we learned

We learned the importance of:

• High-quality data curation for better retrieval results.
• Effective prompt engineering to improve the output of large language models.
• Leveraging RAG and Mistral LLM to build scalable AI solutions.

What's next for Gamerag

Moving forward, we plan to:

• Expand the knowledge base by incorporating more game files and developer input.
• Add customization options to tailor the wizard's responses to specific player preferences.
• Optimize performance for faster query processing and smoother integration with games.
• Explore voice-enabled AI to make the wizard more interactive and engaging for players.

This is just the beginning—we’re excited to see how this tool evolves and enhances the gaming experience! 🎮✨

Built With

• llm
• mistril
• snowflake