Gamepal | Devpost

Step 2: Start a conversation
Step 1: create an account
Step 3: Speak naturally, describe your issue
Step 4: Agent gives you tips and you can enjoy your game
Supports all type of games and consoles

Inspiration

I am a kid from the 90s. I have spent hours on my computer playing games like Doom and Duke Nukem, as well as puzzle games like Myst.

More often than I would like to admit, I got stuck in games and could not make any progress. This is where I would go to the newspaper stand and buy video game magazines. Those magazines they had pages and pages of walkthroughs, describing every single quest and all the different cheat codes to progress faster in the game.

This was my entry into programming: finding the console in The Sims, entering rosebud, and then opening a world of modding. 🤩

Before the magazines, there were hotlines you could call, and real video game experts on the other side would give you hints on how to progress.

I wanted to recreate this so today's gamers could experience a similar thrill while they play.

What it does

Gamepal is an agent specifically trained on walkthroughs available online, such as the amazing Zeldadungeon for all Zelda games.

You can ask questions about the game you are currently playing, such as:

What is the recipe for a stamina elixir?
Where can I find the Chaas Qeta shrine?
Is there a cheat code to obtain full hearts?

If Gamepal knows the answer, it will reference its knowledge base; otherwise, it will search the internet to find it.

How we built it

The app UI has been built using Lovable

The agent interaction is built on top of Elevenlabs Conversational AI.

Knowledge comes from crawling via Firecrawl or on-demand search via a make scenario that is calling Exa API

Finally, I used Supabase for all the user authentication and databases need.

Challenges we ran into

Elevenlabs:

prompting the agent correctly to trigger the client's tools

Accomplishments that we're proud of

Opened PR on Elevenlabs doc

Built something visually appealing

discovered new tools like Exa and Firecrawl

have a better understanding of conversation agent

What we learned

How to create and guide apps with Lovable

How to build a conversational agent with Elevenlabs

How to crawl and search using Exa and Firecrawl

What's next for Gamepal

If I had more time, I could have connected it to a real Twilio number, so really re-enact the original experience. Right now it's only connected to a test number

Also, I would love to build a way to augment the knowledge base automatically so it does not have to search the web.