InfiniteAI

Inspiration

Almost every LLM UI out there present conversations in discrete, unrelated, and linear formats. It is difficult, if not impossible, to see where things are branched off, where models are switched, etc. To find a branch node, and create a new thread of conversation from there. To use and flip flop between models provided by OpenRouter, HuggingFace, Amazon Bedrock, Google Vertex AI, OpenAI, Claude, and other providers.

What it does

It makes all these things possible.

How we built it

Backend was built using Hono, by exposing a REST interface. Front-end was built using D3, Dagre, Pixi and a number of other libraries.

Challenges we ran into

Graph layout and drawing algorithms. Managing an infinite scene graph.

Accomplishments that we're proud of

A working Proof of Concept.

What we learned

Graph drawing algorithms, in particular those that are useful for our use case.

What's next for Infinite Mind

Completing the project, and releasing an MVP/Beta version to the public.

Built With

• claude
• d3.js
• dagre
• deepseek
• hono
• jotai
• next
• openai
• openrouter
• pixi
• supabase
• typescript
• vercel
• vertex