Skip to content

Latest commit

 

History

History
 
 

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 

README.md

Beautiful Chat

What This Demo Shows

A polished, brand-themed chat surface built on top of the shared Claude Agent SDK backend. It demonstrates how far you can take the look-and-feel of a CopilotKit chat with pure frontend dressing — no extra runtime config, no special agent, no declarative-generative-ui catalog.

How to Interact

  • Pick one of the seeded suggestion pills ("Plan a 3-day Tokyo trip", "Explain RAG like I'm 12", "Draft a launch email"), or type your own prompt.
  • The right-side panel is a static decorative read-out (a sparkline + intent breakdown). It does not react to chat traffic — it's there to show how a flagship chat cell can sit alongside ambient app UI.

Technical Details

  • Runtime: this cell uses the shared /api/copilotkit endpoint. The agent ID beautiful-chat is registered alongside the other shared-agent names (agentic_chat, tool-rendering, etc.) in src/app/api/copilotkit/route.ts.
  • Backend: the same default Claude agent the agentic-chat cell uses (src/agents/agent.py). No bespoke prompt or tools.
  • Cosmetic layer: lives entirely in page.tsx — gradient background, brand fonts (Inter), seeded useConfigureSuggestions, and a hand-rolled SVG side panel.

Simplified vs Canonical

The langgraph-python version of this demo ships a much larger surface — ExampleCanvas, useGenerativeUIExamples, an A2UI declarative-generative-ui catalog of charts and cards that the agent can drive directly. Those features depend on streaming-structured-output primitives (A2UI tool envelope, dynamic schema injection) that the claude-sdk-python integration does not currently expose to the showcase.

This port keeps the polished chat shell and treats the dynamic canvas as out-of-scope. If you want the full version, see the langgraph-python beautiful-chat cell.

Building With This

If you're building a real product chat surface, this is a good template:

  1. Pick the agent backend you actually want.
  2. Add a layout-level wrapper around <CopilotChat /> with your brand background, max-width, and rounded card chrome.
  3. Seed useConfigureSuggestions with the prompts that match your domain.
  4. Decorate the rest of the page with whatever ambient UI makes sense — the chat surface is fully self-contained and won't fight your layout.