pyspoor 0.8.20

Python bindings for the deterministic spoor document engine

pyspoor

Typed Python adapter for spoor-core.

from spoor import parse_path

result = parse_path("report.pdf")
print(result.content.value.markdown)
for warning in result.warnings:
    print(warning["code"], warning.get("location"))

Agents must inspect result.warnings; a successful parse can still report a missing PDF text layer, suspicious text, merged-table degradation, or omitted visuals.

Tables: narrowing and pagination

For CSV/XLSX, parse_bytes / parse_path accept the same narrowing options as the CLI, so callers can page through full tables instead of the default 100-row preview:

from spoor import parse_bytes

# A slice by inclusive 1-based row range (mutually exclusive with limit/offset)
parse_bytes(data, source_name="data.xlsx", sheet="Sheet1", rows=(5, 104))

# Or paginate by limit/offset and keep only some columns
parse_bytes(data, source_name="data.xlsx", columns=["分类", "金额"], limit=100, offset=200)

PDFs: page ranges

For page-oriented PDFs, pass pages (inclusive 1-based) to parse only a slice and avoid reading a large document end to end:

parse_path("report.pdf", pages=(1, 3))  # only pages 1–3

Extracting embedded media

Resolve a safe media URI emitted in the output (DOCX images, extractable PDF images) to raw bytes for handing to an external vision model:

from spoor import extract_media

image = extract_media(data, "spoor://docx/part/word/media/image1.png", source_name="report.docx")

Build locally with maturin develop.