SteamPulse is a website that allows users to search for Steam games and view AI powered analysis of the last few hundred reviews for that game. The AI then summarizes the Pros and Cons of that game to make it easier to see if a game is worth buying.
I built SteamPulse because I wanted to build a community in a space that's been important to me all my life, video games. I saw an opportunity to use Steam's Web API along with ChatGPT to get some cool things to happen. Review Sentiment Analysis is the first thing I've done but there's a lot more coming down the pipeline related to
- Roasting user's based on their recently played games
- AI Game recommendations
- Cool analytics and stats on steam games
Please try out what I've built so far and give feedback. It would help a ton!
This is a really cool tool. 2 Questions:
1) Do you cache the results for similar requests? Because feeding all the reviews into the API is for sure not totally cheap, right?
2) In general, how do you reduce the token amount for this? Or don't you reduce at all?
Maybe something you could have a look into is "SPR", it's a technique to do semantic compression. GPT-3 is really good at this and it's a good way to reduce the token amount for gpt-4 models, while still keeping most of the meaning!
Good idea! Congrats on your launch!
Are there plans for the future to add PS Store and Xbox Games reviews as well and create a universal AI-powered game review platform?