Command Line Interface¶
Sci-Fi Parser installs several command-line entry points via
[project.scripts] in pyproject.toml.
Parsing PDFs¶
The main entry point for the package is:
parse-folder INPUT
where INPUT is either a PDF file or a directory containing PDF files.
By default, the parser writes its outputs to an output directory and runs
the full pipeline:
Image extraction
Chart classification
OCR
VLM extraction
Example:
parse-folder train_data/small_pdfs
Custom output directory¶
Use --output (or -o) to specify a custom output directory:
parse-folder train_data/small_pdfs --output results
Disabling pipeline stages¶
Disable chart classification:
parse-folder train_data/small_pdfs --no-classify
Disable OCR:
parse-folder train_data/small_pdfs --no-ocr
Disable VLM extraction:
parse-folder train_data/small_pdfs --no-vlm
Multiple stages may be disabled simultaneously:
parse-folder train_data/small_pdfs --no-classify --no-ocr
Displaying help¶
To see all available options:
parse-folder --help
Other command-line tools¶
Sci-Fi Parser also installs utilities for benchmarking and synthetic dataset generation:
bench-pipeline
synthetic-bars
The benchmark package has its own documentation in
src/sci_fi_parser/benchmark/README.md
which describes the benchmarking workflow, configuration, and output formats in detail.