I wanted to share a project I just launched on Vercel called FindThaGame. It is a discovery tool designed to help you find video games based on fuzzy memories and vague details.
It is built to help people track down those titles that are stuck in your head but hard to find via standard search. It uses a custom scoring engine to rank results by relevance instead of just using basic filters.
The Stack
- Frontend: React 19, Vite, TypeScript
- UI: Tailwind CSS, Shadcn UI
- Deployment: Vercel (Serverless Functions)
- APIs: IGDB (Game data) and Groq AI (Real-time summaries/translations)
Why I built this
Standard search engines usually fail if you don’t know the exact title. I wanted a tool that could handle descriptions like ‘talking fish’ and actually return the correct game.
How it works
The core of the app is a Hybrid Multi-Query Search. When a user types a query, it triggers two parallel processes:
- Strict Query: Searches for full phrases in titles and alternative names.
- Broad Query: Scans individual keywords across summaries, storylines, and metadata tags.
Smart Scoring Logic
To ensure accuracy, I developed a custom scoring algorithm. Instead of filtering games out, it adds or subtracts points based on specific criteria:
- Match Quality: Exact title matches get a +3.0 boost, while keyword hits in the context (summary/storyline) get +0.3.
- Weighted Multipliers: I use multipliers to prioritize main titles over clutter. Main games have a 1.0x multiplier, while DLCs and soundtracks are penalized with a 0.5x multiplier.
- Developer Boost: If the search matches a known developer name, that entry receives a 2.2x multiplier to its final score.
Performance on Vercel
To keep things fast on a serverless architecture, I implemented server-side token caching for the IGDB API and a warmup middleware to reduce cold start latency.
Live Site: https://findthagame.vercel.app/
GitHub Repo: https://github.com/kbtale/findthagame
I would love to get your thoughts on the scoring logic or the search performance!
