T2Chat

• 

  T2Chat

Inspiration

One of my favourite tech creators made a project called 'T3Chat', which was really well liked by us fans. He recently had started a competition about 'who can make the best OS T3Chat, which is even better than the original' and me who was already participating in the bolt hackathon knew, that he was also a judge here, and that's how I thought I should make this for both their challenges.

What it does

T2Chat is an Open Source LLM Wrapper, and you can use it to use any LLM of your choice, with your own API keys, or the in-house free models. There are all features from all AI providers, all condensed into one single website. There are features like AI voice calls, realtime device sync, web search, image generation, image and pdf analysis, TTS+STT, BYOK, UI Customisation, and so much more that it can't be explained all at once, most of all it has a very premium UI/UX.

How we built it

Initially prototyped structure with Bolt, till the best it could do, once we ran into a dead end we downloaded the codebase and went into our IDEs, there we did all the work needed ourselves, and the implementation it needed. The website is built on top of NextJS, Tailwind, Typescript, for auth we've used better-auth + postgres, backend and db is all done on Convex, real time sync done using Dexie. Tools used have been netlify/vercel/v0/cursor/claude

Challenges we ran into

Setting up convex with better-auth was a major difficulty, improving the website performance was a big challenge as well, and lastly tweaking the UI to ensure the best UX was a challenging yet fun job. All these done using mostly AI was a challenging task as it couldn't do exactly what we wanted.

Accomplishments that we're proud of

We managed to make a really good looking brand identity by the theme and design we used, and the marketing video we released! it was liked very well by people on Twitter.

We made out product feature packed and added as many new things as we could, and proudly I can say we achieved our goal.

What we learned

Use of AI in making things is very helpful, but only if you have proper knowledge, otherwise you'll be stuck on things the AI has trouble solving.

Always keep trying out new technologies, it works out best in the long run.

What's next for T2Chat

We want to make T2Chat a product widely loved by people, and hope they are willing to contribute to making it the best one out there! Our team, though only consisting of 2 people, will keep making things better, we will keep adding more features and improvements regardless of us winning or losing the hackathon, this is our own project built with love and hardwork more than something limited to a hackathon.

Built With

• better-auth
• convex
• dexie
• nextjs
• postgresql
• typescript