Skip to main content

Live Chat Demo

The embedded demo is the same Flutter example app from example/lib/main.dart, built as a static web app and served from the documentation site.

This page is refreshed for the 0.3.7 release so the docs demo matches the current shipped example. The high-resolution recording uses only deterministic offline fixtures: it demonstrates flat source-backed selection through formatted text and inline code, the semantic transition from an unfinished inline link to a completed link, and character-level selection inside a custom widget while its switch remains interactive.

The embedded app also includes the default Fade reveal, cross-table dragging, TextField-like scroll-edge selection, LaTeX rendering, and web parser support. Completed assistant renderers retain their streaming state when new messages are added. Use the link and pointer icons in the app bar to open the two offline labs without configuring a chat provider.

It can connect to:

  • local Ollama through http://localhost:11434
  • OpenAI / ChatGPT compatible streaming APIs
  • Anthropic Claude Messages streaming
  • Gemini streamGenerateContent
  • xAI Grok through the OpenAI-compatible chat completions API

Local Ollama​

Run Ollama locally, pull a model, then use the default provider in the demo:

ollama serve
ollama pull llama3.2

The demo sends requests to /api/chat and renders streamed Markdown responses with AnimatedStreamingMarkdown.

Cloud Providers​

Pick a provider from the dropdown, enter a model and API key, and send a message. API keys entered in the demo stay in the browser session. The docs build does not bundle .env secrets.