Conversation-Tree

Conversation-Tree

• One-line description: An AI chat interface that explores conversations not as a single linear record but as an infinitely branching tree.
• Period: 2025.06 – 2026.02
• Role: Personal Project (Planning, Frontend Development)
• Tech Stack: React 19, TypeScript, D3.js, Tailwind CSS, Gemini API, Vite

Reason for Creation / Problem

AI typically provides a lot of information at once, often in a chatroom format. For example, if explanations were divided into points 1, 2, and 3, it was cumbersome to delve into point 1 and then return to point 2. I created this project with the idea that it would be better to separate conversations into tracks and continue them only within the desired context.

Key Features

• Multiverse Branching: Create new branches from any message, with each branch maintaining an independent context.
• Memory Linking: Inject context between different conversation paths to cross-reference ideas.
• Universe Map: Visualize the entire conversation structure in real-time with D3.js, and explore by dragging and zooming nodes.
• Track Comparison: Select multiple conversation paths simultaneously for AI to analyze and compare side-by-side.

My Own Considerations

The starting point was the inconvenience that when AI pours out information all at once, for example, if it explains things divided into points 1, 2, and 3, it's difficult to delve into point 1 and then return to point 2. Expanding on the idea of the "Hierarchical Semantic Memory System" I previously created to compensate for AI's lack of long-term memory, I conceived the idea of separating memories into tracks and continuing conversations only within the desired context. However, separating conversations led to the problem of context being disconnected between different tracks. To address this, I designed both memory linking, which injects context between tracks, and branch referencing, which refers to other branches.

Improvement program is under development. (Nodi Project)