> ## 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.

# Import From Contracts

> Turn contracts you already negotiated into a starting playbook, including consensus across a whole folder of them

The fastest honest answer to "what is our position on liability?" is usually sitting in the contracts you already signed.
Import reads those contracts, splits them into clauses, and infers a starting position for each one.
You review the result and edit it before anything is saved.

<Note>
  Import never saves a playbook on its own.
  It produces a preview you can edit, drop rows from, and then save through the normal create flow.
</Note>

## Start an import

From the new-playbook page, choose **Import from a document**.

<Steps>
  <Step title="Set the contract type">
    Leave it on **Auto-detect from document** and Vaquill infers the type from the text, or pick one explicitly if you already know.
  </Step>

  <Step title="Give it the contract">
    **Upload file** takes Word and PDF files.
    **Paste text** takes the body of a contract pasted directly, which is useful when you only have a fragment or an email attachment you cannot download.
  </Step>

  <Step title="Extract">
    One file or pasted text runs a single extraction.
    Two or more files run extraction on each and then consolidate them into one consensus playbook.
    The same button does both; the file count decides the mode.
  </Step>

  <Step title="Review the preview">
    Every recognized clause appears as an editable row: preferred position, acceptable range, escalation triggers, and rationale.
    Uncheck any row you do not want. Fix anything that reads wrong.
  </Step>

  <Step title="Name it and save">
    The preview seeds the normal playbook form. Add a name, confirm the contract type, and create.
  </Step>
</Steps>

## One contract, or many

<Tabs>
  <Tab title="A single contract">
    Best when you have one agreement that genuinely represents your house position, for example the template your team actually sends.

    Each clause block is classified and summarized into a preferred position, an acceptable range, escalation triggers, a rationale, and, where the clause has typed constraints, the numbers themselves: cap, cure period, payment days, and so on.

    Repeated clause types are merged rather than duplicated.
  </Tab>

  <Tab title="Several contracts at once">
    Best when the right answer is the consensus across a body of signed deals, not whichever one you happened to pick first.

    Every file is extracted, then the results are consolidated clause by clause.
    Each consolidated position carries a source count so you can see how many of your contracts agreed, and non-consensus alternates are kept as minority positions rather than discarded.

    Files that produce no recognizable clauses are skipped with a notice, and you need at least two files that did produce clauses for the consolidation to run.
  </Tab>
</Tabs>

## What it accepts

| Input            | Limit                                               |
| ---------------- | --------------------------------------------------- |
| Word files       | 25 MB each                                          |
| PDF files        | 10 MB each, and the PDF must have a real text layer |
| Files per import | 20                                                  |
| Pasted text      | 400 characters minimum, 500,000 characters maximum  |

<Warning>
  A scanned PDF with no text layer is rejected with a message telling you to run OCR first.
  Convert it through the OCR tool in [Legal Tools](/docs/guides/legal-tools), then import the result.
</Warning>

## Resume a recent import

The new-playbook page lists your recent extractions.
**Resume** reopens one of them in the preview stage, so an interrupted session does not mean re-uploading and re-running the extraction.

## After the import

An imported playbook is a first draft assembled by a model from your own paper.
Two things are worth doing before you rely on it.

* Run the [structure check](/docs/guides/playbooks/structure-check). Imports typically arrive with language and no governance, so the score starts low and the findings tell you exactly what is missing.
* Add fallback ladders and sign-off levels to the clauses you actually negotiate. Extraction infers what your contracts say, not how far you were willing to move or who had to approve it.

## Limitations

* Clause detection relies on document structure: numbered headings, all-caps headings, or a bold first line. A contract with unusual formatting produces fewer, coarser blocks.
* Very large contracts are analyzed up to a cap. When that happens you get a warning that the analysis was partial, and any missing clauses have to be added by hand.
* Clause classification is model-judged. Near-miss labels are folded onto the standard clause type, and clause types outside the standard vocabulary are kept only up to a small number per document.
* Consolidation across several contracts is model-judged too. Confidence is a coarse measure of how many source files agreed on a clause, not a similarity score on the text.
* Extraction reads the final language of a contract. It does not learn from tracked changes, so it tells you where you landed, not what you pushed for.
* Positions arrive without fallback ladders, walk-away floors, approval levels, or conditional rules. Those are yours to add.
* Nothing is verified against law. An imported position reflects what your contract said, including any mistakes in it.

## Related

<CardGroup cols={2}>
  <Card title="Building a Playbook" icon="pen-ruler" href="/docs/guides/playbooks/building-a-playbook">
    Edit and extend the imported result.
  </Card>

  <Card title="Structure Check" icon="clipboard-check" href="/docs/guides/playbooks/structure-check">
    See exactly what an import is missing.
  </Card>

  <Card title="Starter Templates" icon="layer-group" href="/docs/guides/playbooks/templates">
    Start from a curated playbook instead, then import selectively.
  </Card>

  <Card title="Supported File Formats" icon="file-lines" href="/docs/guides/file-formats">
    What Vaquill can read across the product.
  </Card>
</CardGroup>