Learning pathsAI Hub fundamentals

Introduction to the flow editor and advanced apps

Learn about advanced apps built with the flow editor and how they differ from standard apps built in the app editor.

Using the flow editor is covered in more detail in the Flow editor and advanced apps course in the Developer learning path.

What is an advanced app?

This video introduces how you can create advanced document automation apps with the flow editor and custom Python code. Learn how advanced apps can support use cases calling for multi-document packets, unstructured tables, multi-stage routing, or logic beyond visual configuration.

Topics include:

  • Designing ingestion, classification, branching, extraction, and checks on the flow canvas.

  • How user-defined functions (UDFs) extend visual steps for setup, per-file processing, aggregation, and export.

  • How the unified product lets teams start in the app editor and extend into advanced workflow building tools without leaving the platform.

Using Flow

Learn about the core steps used when building flows, then plan a document pipeline on the flow canvas that uses the following steps:

  • Process files — Digitize raw files with OCR and image pre-processing. Optional JSON settings can tune how the text layer is built.

  • Agent classifier — Categorize files from layout and text using natural language class definitions.

  • Filter — Route classified files down parallel branches so multi-document packages stay organized.

  • Agent extract — Extract fields and complex tables with generative AI, guided by natural language prompts.

  • Combine — Merge parallel branches back into a single output packet.

  • Apply checkpoint — Run validation rules as a quality gate. Failing records can pause for human-in-the-loop review.

Using custom UDFs

Extend visual flows with Python user-defined functions (UDFs). Learn how to create a UDF module from the modules panel, generate the scripts folder, then edit scripts.py in the built-in editor.

There are four primary ways to run custom code in a flow:

  • Pre-flow UDFs — Run before execution starts; use for setup such as copying batches from external systems.

  • Map UDFs — Process individual files or records in parallel.

  • Reduce UDFs — Aggregate results across multiple outputs into a consolidated batch summary.

  • Post-flow UDFs — Run after execution finishes; use for cleanup, reporting, or sending results to downstream systems.

Pre-flow and post-flow hooks live in the events panel. Map and reduce UDFs attach as steps on the canvas. You can also call UDFs from Refiner and register custom validation logic in validations.py.