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

# Calibrate scenarios

> Test the Book against generated boundary cases and historical quotes before it can reach customers.

Check representative and historical situations before a rule can reach customers. Calibrate confirms that the Book produces the right answer for cases that matter.

## Before you start

* A workspace with an active Test environment
* A draft exists and can be evaluated (extraction has completed and the draft compiles)

## What the Calibrate page shows

The page has three sections:

1. **Generated tests** — Kordless chooses boundary situations where a small detail may change the answer. You decide the expected outcome.
2. **Historical cases** — Decide whether a past quote is a rule to preserve or a one-off mistake that should not govern new work.
3. **Regression suite** — Confirmed cases that replay every time the Book is compiled or published.

## Generated tests

Kordless generates boundary cases — situations where the details matter and a small change might produce a different result. Each generated case shows:

* A label describing the case
* The inputs — the specific values being tested
* The current result — what the Book currently produces for these inputs
* The statements the result was derived from

You decide the expected outcome for each case using four buttons:

| Button                          | What it means                                                                                                          |
| ------------------------------- | ---------------------------------------------------------------------------------------------------------------------- |
| Correct                         | The current result matches what you expect. The case is confirmed and added to the regression suite.                   |
| Wrong price                     | The current result does not match. You can enter the expected price to record a correction.                            |
| It depends                      | The result should go to review rather than being quoted automatically. The case is recorded as **Needs owner review**. |
| I would not quote this remotely | The work cannot be responsibly priced remotely. The case is recorded as **Inspection required**.                       |

<Note>
  A correction produces a proposal — it does not mutate Live. The proposed change appears on the [Preview and publish](/product/book/preview-and-publish) page for owner review.
</Note>

If a generated case has missing inputs or references an unsupported rule, you cannot confirm it yet. The page tells you to resolve the missing inputs or unsupported rule first.

## Historical cases

When you import pricing from a source that includes past quotes, Kordless may import them as historical calibration cases. Each historical case shows:

* The label describing the case
* The inputs
* How it was imported — as a priced result or another outcome

You decide whether each historical case should be:

* **Preserved** — the result is correct and should govern new work. The case is added to the regression suite.
* **Corrected** — the result was a one-off or mistake and should not govern new work. You can enter the expected price.

This lets you separate rules you want to keep from exceptions you do not want to repeat.

## Regression suite

Confirmed cases — both generated and historical — become part of the permanent regression suite. Every case in the suite replays each time the Book is compiled or published.

Each regression case shows:

* The label
* The source (generated or historical)
* The offering it applies to
* A status chip: **Passing**, **Needs attention**, or **Not run**

A case marked **Needs attention** means the current Book draft produces a different result from what you confirmed. This is a signal that something changed and needs your review before publishing.

## How calibration affects the Book state

The Book overview shows **Needs calibration** when cases are waiting for a decision. Once you confirm or correct every pending case, the Book state advances to the next stage — typically **Needs verification**.

You can return to Calibrate at any time to review or add cases. New generated cases may appear when sources change or new rules are added.

## If it does not work

<Accordion>
  <AccordionItem title="Nothing to calibrate yet">
    If the page shows "Nothing to calibrate yet," add pricing and review the draft first. Generated and historical cases appear here when the Book has something to test.
  </AccordionItem>

  <AccordionItem title="Cannot confirm a generated case">
    If a generated case shows "Resolve the missing inputs or unsupported rule before confirming this case," go to Review draft or Answer questions to resolve the underlying issue first.
  </AccordionItem>

  <AccordionItem title="Permission-limited member">
    Members can view the Calibrate page but may not be able to record decisions depending on their role. Ask a workspace administrator if you need elevated access.
  </AccordionItem>
</Accordion>

## Next

<CardGroup cols={2}>
  <Card title="Verify Statements" icon="shield-check" href="/product/book/verify">
    Approve material rules through plain-language review.
  </Card>

  <Card title="Preview and publish" icon="upload" href="/product/book/preview-and-publish">
    Review consequences and publish an immutable Book version.
  </Card>
</CardGroup>