The Parse phase
Parse is the stage that reads a document's structure before anything is split, classified, or extracted — layout, language, and text — and sets the output format every later stage works from.
What Parse owns
Parse turns a normalized document into a structured reading of the page: where the regions are, what language the text is in, and what the text says. It runs after Initialization and before Split / Classify, and every stage downstream — Split, Classify, Extract — works from what Parse produces. Parse is an umbrella phase: it has no logic of its own, it groups a small set of sub-stages that each answer one question.
The Parse stage selected in the pipeline builder. The left rail is the stage order; the right panel is the Parse stage's own configuration, open on its three sub-stages.
The Parse sub-stages
Parse carries three sub-stages, each independently toggled and each bound to an engine and a confidence threshold.
| Sub-stage | Answers | Typical engine |
|---|---|---|
| Layout detection | Where are the regions — blocks, tables, figures — on each page? | Azure Document Intelligence — Layout |
| Language detection | What language is the document written in? | LangDetect |
| OCR | What is the text, and where does each token sit on the page? | Azure Document Intelligence — Read |
The counter on the stage card (3 of 3 sub-stages on) reflects how many are enabled. A sub-stage that is turned off is skipped for every document the pipeline runs.
Output
Parse decides the shape of the document handed to later stages.
- Parse output format — Markdown preserves structure (headings, tables, lists); Text is a flat transcription. The choice is what Split, Classify, and Extract read.
- Include bounding boxes — pixel coordinates for every detected element. Required for visual review, where a reviewer sees the value highlighted on the page.
- Include confidence scores — per-element scores. These are the signal that downstream gating and Reviews routing key off.
Chunking
For documents larger than a model can read in one pass, Parse can emit the document in chunks rather than whole. Chunking is off by default; when on, the pipeline passes chunks to later stages under the strategy configured on the stage.
How a stage is configured
Every stage and sub-stage in a pipeline shares the same controls: an enabled toggle, an engine, a model where the engine is an LLM, and a confidence threshold. Entry and exit webhooks can fire around any stage. A pipeline whose stage order is invalid — for example Extract with nothing to classify against — is rejected before any document runs, so a published pipeline is always one that can execute.