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

# Schedules & custom rules

> Run scans automatically on a cadence and extend the detection patterns engines use.

Two related ways to keep the platform working for you without manual
clicks: scheduled scans and custom detection rules.

## Scheduled scans

Catch drift and new issues as soon as they appear. A schedule re‑runs
one or more engines on a cron expression, against either a dataset or
every dataset in a project.

### Where to set them up

| From                               | Result                                                               |
| ---------------------------------- | -------------------------------------------------------------------- |
| **Dataset detail → Scheduler**     | Schedule scoped to this dataset only.                                |
| **Project detail → Schedules**     | Schedule that fans out to every dataset in the project in one sweep. |
| **Settings → General → Schedules** | Workspace‑wide view to list, edit, or disable any schedule.          |

### What you configure

| Field                   | Notes                                                                                   |
| ----------------------- | --------------------------------------------------------------------------------------- |
| **Cron expression**     | UI provides presets (hourly, daily 02:00, weekly Monday, ...) or type your own.         |
| **Engines**             | One or many. They run in parallel as worker capacity allows.                            |
| **Parameter overrides** | Same controls as a manual scan (sample cap, GPU, advanced parameters).                  |
| **Notifications**       | Email and / or webhook on completion. See [Integrations](/data-integrity/integrations). |
| **Concurrency policy**  | Auto‑cancel if the previous run is still going, or let them stack.                      |

### Where you see them run

Scheduled runs appear in the scan list with a clock icon and the
schedule ID in the metadata column. Failures surface in:

* The Dashboard's **Recent activity** widget.
* The notification channel you configured.
* The audit trail (with parameter snapshot).

<Tip>
  Use **Auto‑cancel previous run** on schedules that run more often
  than scans typically finish. Stacked queued scans waste capacity
  without giving you fresher data.
</Tip>

## Custom rules

Custom rules extend the detection patterns the engines use. Today
they're consumed by **Text Analysis** PII / secret detection;
additional engines pick them up as the feature expands.

### Rule structure

| Field        | What it controls                                             |
| ------------ | ------------------------------------------------------------ |
| **Name**     | Human label, e.g. `Internal employee ID`.                    |
| **Pattern**  | Regex. Validated for catastrophic backtracking before save.  |
| **Severity** | `CRITICAL`, `MEDIUM`, or `INFO`. Drives the finding's badge. |
| **Score**    | Base score applied when the rule matches (0–1).              |
| **Scope**    | Global, per‑project, or per‑dataset.                         |
| **Category** | Free‑form, typical values: `pii`, `secret`, `phi`, `custom`. |
| **Enabled**  | Toggle without deleting.                                     |

### Where to manage them

**Settings → Custom Rules.** The page lists every rule with its
scope, category, severity, and enabled state. From here you can:

* **Create** a new rule.
* **Edit** an existing rule's pattern, score, or scope.
* **Archive** to hide from active scans without deleting.
* **Delete** outright.
* **Bulk import / export** rules as JSON. Useful for moving rules
  between workspaces.

### Testing before you save

Every rule editor has a **Test** button. Paste sample text; Blindsight
runs the pattern against it and highlights matches in real time. Use
this to confirm you're not over‑matching before pushing the rule
live.

<Steps>
  <Step title="Draft the rule">
    Pick a name. Write the regex.
  </Step>

  <Step title="Test it">
    Paste sample text that should match and text that should not.
    Tweak until both are correct.
  </Step>

  <Step title="Pick a scope">
    `Global` for org‑wide patterns. `Project` or `Dataset` when the
    rule only makes sense in context.
  </Step>

  <Step title="Save and confirm">
    The rule becomes live within \~30 seconds (a small cache TTL).
    Next text scan will use it.
  </Step>
</Steps>

### Common patterns

<AccordionGroup>
  <Accordion title="Internal identifiers">
    Most companies have a structured ID format (`EMP-123456`,
    `CUST-AB1234`, `INV-2025-…`). Capture each as a rule with
    severity `MEDIUM` and category `pii`.
  </Accordion>

  <Accordion title="Proprietary phrases">
    Code names for unreleased products, M\&A targets, internal project
    names. Severity `CRITICAL`, category `secret`, score 0.9.
  </Accordion>

  <Accordion title="Regulatory identifiers">
    Industry‑specific identifiers (Swiss AHV, US SSN, EU VAT, ICD‑10
    codes). Use one rule per format, scope `Global`, severity
    `CRITICAL`.
  </Accordion>
</AccordionGroup>

<Warning>
  Catastrophic‑backtracking guard is on by default. A pattern that
  could blow up at scale is rejected before save with a clear error.
  If a rule is rejected, simplify or anchor the pattern.
</Warning>