wehub-resource-sync
34 lines
1.1 KiB
Markdown
34 lines
1.1 KiB
Markdown
# QA: Open-Ended Generative UI — Claude Agent SDK (Python)
|
|
|
|
## Prerequisites
|
|
|
|
- Demo is deployed and accessible
|
|
- Agent backend is healthy
|
|
- ANTHROPIC_API_KEY is set
|
|
|
|
## Test Steps
|
|
|
|
### 1. Basic Functionality
|
|
|
|
- [ ] Navigate to `/demos/open-gen-ui`
|
|
- [ ] Verify the chat renders with minimal suggestions visible
|
|
- [ ] Click the "3D axis visualization (model airplane)" suggestion
|
|
- [ ] Verify a sandboxed iframe appears with the generated visualization
|
|
|
|
### 2. Feature-Specific Checks
|
|
|
|
- [ ] Verify the iframe contains SVG-based content (not div-stacks)
|
|
- [ ] Try the "Quicksort visualization" suggestion — verify an animated bar sort renders
|
|
- [ ] Verify the visualization is self-running (no user interaction needed)
|
|
|
|
### 3. Error Handling
|
|
|
|
- [ ] Verify no CORS/CSP errors in the console
|
|
- [ ] Verify the sandbox iframe renders within the chat turn
|
|
|
|
## Expected Results
|
|
|
|
- Agent calls `generateSandboxedUi` once per turn
|
|
- The runtime's OpenGenerativeUIMiddleware converts the call into activity events
|
|
- The built-in renderer mounts the agent-authored HTML inside a sandbox
|