# Release Summary: API Reference Book

- Source: https://threejs.org/manual/en/creating-a-scene
- Version: n/a
- Output format: both
- Public only: no
- Run health: **🔴 shell-page storm**

## Build
- Pages in PDF/HTML book: 64
- Chapters: 2

## Crawl
- URLs discovered: n/a
- URLs crawled: 1
- Crawl success: 1/1 (100.0%)
- Truncated at 50,000 chars: 0
- Shell pages (SPA shells): 1

## Extraction Quality
- Successful LLM extractions: 1
- Fallback pages (HTML parser): 63 (98.4%)

### Failed Extraction Details (first 5)
- **https://threejs.org/manual/en/animation-system**: Finish Reason = `tool_calls`, Error = `JSON parsing failed`
- **https://threejs.org/manual/en/backgrounds**: Finish Reason = `tool_calls`, Error = `JSON parsing failed`
- **https://threejs.org/manual/en/billboards**: Finish Reason = `tool_calls`, Error = `JSON parsing failed`
- **https://threejs.org/manual/en/cameras**: Finish Reason = `tool_calls`, Error = `JSON parsing failed`
- **https://threejs.org/manual/en/canvas-textures**: Finish Reason = `tool_calls`, Error = `JSON parsing failed`
- … and 58 more (run with `--verbose` or open `extracted_pages.json`)

## Self-Contained Output
- Inline images: enabled (threshold: 2000000 bytes)
- Images inlined as `data:` URIs: see HTML artifacts
- Images written to disk: 0

## Artifacts
- PDF: `api-reference-book.pdf`
- HTML: `api-reference-book.html`
- Manifest: `release-manifest.json`

## Page Types
- reference: 63
- guide: 1

## Notes
- >50% of pages fell back to HTML parsing (98.4%).
- 1/1 raw pages were SPA shells (iframe-bypass may not be enabled).