> ## Documentation Index
> Fetch the complete documentation index at: https://vaquill.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Extraction steps

> Summarize a document, pull a dated chronology, or extract structured contract facts for scoring

Three steps, all AI-kind. They turn document text into something structured that downstream steps can act on.

***

## Summarize Document

<CardGroup cols={2}>
  <Card title="Kind" icon="robot">AI</Card>
  <Card title="Latency" icon="hourglass">Slow</Card>
</CardGroup>

Generates a structured summary of one document with verified citations. The markdown output feeds straight into Export DOCX, Verify Citations, or a downstream agentic step.

**Config**

| Field          | Type            | Required | Notes                                                     |
| -------------- | --------------- | -------- | --------------------------------------------------------- |
| Document       | Document picker | Yes      | One document.                                             |
| Summary length | Select          | No       | Longer summaries include more detail. Defaults to medium. |

**Outputs**

| Field                         | Contents                                                          |
| ----------------------------- | ----------------------------------------------------------------- |
| Document Id, Title            | Which document was summarized.                                    |
| Summary Markdown              | The summary body. Wire this into Export DOCX or Verify Citations. |
| Section Count, Citation Count | Structure and grounding counts.                                   |

<Tip>
  To summarize a whole set rather than one document, put this step inside a [Loop Over Documents](/docs/workflows/builder/steps/routing). That is exactly what the "Summarize Each Document" starter template does.
</Tip>

***

## Extract Chronology

<CardGroup cols={2}>
  <Card title="Kind" icon="robot">AI</Card>
  <Card title="Latency" icon="hourglass">Slow</Card>
</CardGroup>

Pulls dated events from one or more documents and merges them into a single timeline. Uses the same extractor as the matter Chronology tab.

**Config**

| Field          | Type          | Required | Default | Notes                                                                                                                                            |
| -------------- | ------------- | -------- | ------- | ------------------------------------------------------------------------------------------------------------------------------------------------ |
| Documents      | Document list | Yes      |         | 1 to 200. **This one field takes many documents**, and the step fans out across all of them with bounded concurrency, so you do not need a loop. |
| Maximum events | Integer       | No       | 100     | 1 to 500 events on the merged timeline.                                                                                                          |

**Outputs**

| Field             | Contents                    |
| ----------------- | --------------------------- |
| Events            | The merged, ordered events. |
| Event Count       | How many survived the cap.  |
| Timeline Markdown | Renderable timeline body.   |

***

## Extract Contract Facts

<CardGroup cols={2}>
  <Card title="Kind" icon="robot">AI</Card>
  <Card title="Latency" icon="gauge">Medium</Card>
</CardGroup>

Reads a contract and extracts the structured facts (term lengths, carve-outs, durations) that Score Against Benchmarks consumes.

**Config**

| Field           | Type                                | Required | Notes                                         |
| --------------- | ----------------------------------- | -------- | --------------------------------------------- |
| Document        | Document picker                     | Yes      | The contract to read.                         |
| Benchmark suite | NDA, MSA, Employment, M\&A, Privacy | Yes      | Which suite the extracted facts should match. |

**Outputs**

| Field                        | Contents                                                                     |
| ---------------------------- | ---------------------------------------------------------------------------- |
| Facts                        | The structured facts dict. Wire this straight into Score Against Benchmarks. |
| Extracted Count, Total Count | How many of the suite's facts were actually found.                           |

<Note>
  Pair **Extract Contract Facts → Score Against Benchmarks** and select the same benchmark suite on both. That gets you a scored contract without typing a single value. A low Extracted Count relative to Total Count is a signal that the contract is unusual or the wrong suite is selected.
</Note>

***

## Related

<CardGroup cols={2}>
  <Card title="Analysis steps" icon="chart-simple" href="/docs/workflows/builder/steps/analysis">
    Scoring, comparing, and synthesizing what you extracted.
  </Card>

  <Card title="Routing steps" icon="code-branch" href="/docs/workflows/builder/steps/routing">
    Fanning a single-document step across a whole set.
  </Card>
</CardGroup>