API reference

The reference describes the functions used to build a Flowde workflow. Function signatures below are generated from the package source. For an introduction with runnable examples, begin with the pipeline guide.

Page	Contents
Pipeline functions	Directory-based extraction, classification, rotation and parsing
Function factories and helpers	Paddle, OpenAI and Gemini factories; connection checks; partial-result joining; custom function settings
Data types	Classification and parsing schemas, few-shot examples, usage reports and token prices
Benchmarks	Benchmark constructors, metric methods and detailed matching results

Pass filesystem paths as pathlib.Path objects. Directory functions inspect top-level files, sort the input paths, and save results in the supplied save_dir. The run-management guide explains the shared on_existing options.

The reference focuses on the workflow entry points and their public helpers. Modules whose names begin with _ implement the saved-state and worker machinery; applications should use the pipeline functions to manage those operations.