Crate ethos_pdf 

ethos-pdf — WS-ENGINE lane (Milestone A)

The only crate that loads PDFium (invariant 3). Everything crossing EthosPdfBackend is already normalized + quantized (invariant 1: quantize-at- extraction lives here); public schemas/APIs never see PDFium types.

This first WS-ENGINE slice uses a small dynamic FFI boundary over the PDFium C API. Runtime loading is explicit through ETHOS_PDFIUM_LIBRARY_PATH, so parser output cannot accidentally depend on an unknown library from a host search path.

Structs

GeometryProbeChar

Per-character geometry probe record.

GeometryProbePage

Per-page debug geometry signals.

GeometryProbeReport

Debug-only report of PDFium text geometry signals.

GeometryProbeRun

Parser-like text run with alternative PDFium geometry sources.

GeometryProbeSymbols

Optional PDFium geometry symbols discovered at runtime.

PdfiumBackend

PDFium backend implementation.

RawCrop

Raw crop rendered from a PDF page.

Constants

PDFIUM_ARTIFACT_PATH_ENV

Optional environment variable containing the downloaded Phase 1 release artifact path.

PDFIUM_LIBRARY_PATH_ENV

Environment variable containing the exact PDFium dynamic library path.

PDFIUM_VERSION_ENV

Optional environment variable carrying the pinned PDFium release/version string.

QUANTUM_PER_POINT

Profile quantization: 100 quanta per PDF point.