When my LOL player friends forced me to stop playing Dota2 for a “better” game, I found myself lost in champions, builds and combos. Overlays like professor were good but very static and have no interaction or reasoning to give me understanding of how the game works. I wanted to know what should be my next move and why!
* Our Solution *
We made a windows app that works in the background, you press a button and ask your question, then our agent responds to you based on the game’s live events and the guides we found on the web.
Thanks to the LLMs in use, we are supporting 28 native languages too!
* How We Did It *
- Windows client that captures Riot’s official API live events
- The agent is Langchain with tool calls fetching champions guides and builds when game starts.
- FastAPI python, Auth0, Traefik and docker as infrastructure.
- Our main model is gemini-flash-lite-latest, we found it a good choice based on multimodal input capability, latency and price.
- Currently responding with about 2s latency on average for each request.
Our plan: we have no plans, just looking for feedback. It’s cheap enough for us to keep it free for now, and we’d like to make it open source the moment it’s a spaghetti-not. On the product side, we will encourage overlays later and also in an interactive way, meaning users can ask a question and see the suggestion in the for of overlay.
Webpage: sensii.gg
Download link: https://api.sensii.gg/api/v1/downloads/sensii.exe
Discord: https://discord.gg/un6QXZ7Prg
When my LOL player friends forced me to stop playing Dota2 for a “better” game, I found myself lost in champions, builds and combos. Overlays like professor were good but very static and have no interaction or reasoning to give me understanding of how the game works. I wanted to know what should be my next move and why!
* Our Solution *
We made a windows app that works in the background, you press a button and ask your question, then our agent responds to you based on the game’s live events and the guides we found on the web. Thanks to the LLMs in use, we are supporting 28 native languages too!
* How We Did It *
- Windows client that captures Riot’s official API live events - The agent is Langchain with tool calls fetching champions guides and builds when game starts. - FastAPI python, Auth0, Traefik and docker as infrastructure. - Our main model is gemini-flash-lite-latest, we found it a good choice based on multimodal input capability, latency and price. - Currently responding with about 2s latency on average for each request.
Our plan: we have no plans, just looking for feedback. It’s cheap enough for us to keep it free for now, and we’d like to make it open source the moment it’s a spaghetti-not. On the product side, we will encourage overlays later and also in an interactive way, meaning users can ask a question and see the suggestion in the for of overlay.
We’d be happy to know what you think of it!