# seleniumbase-website-scraper

Browser-to-API discovery agent.

This upgrades the original SeleniumBase website scraper idea into a hosted API
discovery workflow:

- opens a site in a browser
- records observable XHR/fetch/document traffic
- captures JSON response bodies when the browser can read them
- filters analytics/static noise
- templates paths and query parameters
- splits GraphQL/JSON-RPC style dispatches
- infers request and response JSON schemas
- emits `openapi.json`, `openapi.yaml`, `report.md`, `index.html`,
  `client.mjs`, `confidence.json`, and redacted samples

## Run Locally

```bash
python -m pip install -r requirements.txt
python -m playwright install chromium
python agent.py https://example.com --wait-seconds 8 --max-clicks 4
```

## Run as A2A

```bash
a2a dev
a2a test --invoke --skill discover_api_from_browser --args-json '{"url":"https://example.com","wait_seconds":8,"max_clicks":2}'
```

Use `origins` to focus on the API host:

```json
{"url":"https://app.example.com","origins":["api.example.com"],"min_samples":2}
```

The replay skill accepts JSONL request/response captures and runs the same
OpenAPI emitter without opening a browser.