Graph Chat

Screen.Recording.2026-10-03.at.5.58.32.PM.mov

An LLM chat where every message is a node in a force-directed graph. Replies spawn new nodes linked to the message before them. Click any earlier node to branch the conversation from there.

Setup

npm install
cp .env.example .env   # then fill in LLM_API_KEY and LLM_MODEL
npm run dev            # API on :8787, UI on http://localhost:5173

.env is gitignored. The browser never sees the key or endpoint: it only calls /api/chat, and the Node server forwards the request to the provider.

Any OpenAI-compatible /chat/completions endpoint works. To switch providers, change LLM_BASE_URL, LLM_API_KEY and LLM_MODEL.

How branching works

• Assistant node selected → your message continues from that reply.
• User node selected → your message becomes an alternative to it (a sibling branch).
• Start node selected → begins a new conversation.
• Regenerate → another reply to the same user message, as a new branch.

Only the path from the start node to the new message is sent to the model as context. Conversations are saved in localStorage.

Production

npm run build && npm start   # serves dist/ and the API from one process