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

# Deployment and visual proof with EyeParity and DIPP

> Ship SpreadKnowledge through GitLab CI/CD with EyeParity vision analysis and DIPP corroboration as the reviewer source of truth.

SpreadKnowledge deploys only through GitLab CI/CD, and reviewers rely on EyeParity's production route as the source of truth. This page describes the deployment path and the visual-proof workflow you must complete before requesting review.

<Warning>
  GitLab CI/CD is the only supported deployment path for production and shared environments. Local deployment shortcuts, manual uploads, and out-of-band pipelines are unsupported and will be rejected in review.
</Warning>

## Deployment path

<Steps>
  <Step title="Open a merge request">
    Push your branch and open a GitLab merge request. CI runs `yarn type-check`, `yarn build`, and `yarn docs:mintlify:check` on every pipeline.
  </Step>

  <Step title="Pass the docs check">
    `yarn docs:mintlify:check` must be green. A failing docs check blocks the pipeline.
  </Step>

  <Step title="Deploy through CI/CD">
    Promotion to dev and prod is performed by the GitLab pipeline. Do not deploy from a workstation.
  </Step>
</Steps>

## Visual proof workflow

<Steps>
  <Step title="Capture production">
    Capture the production route for every surface your change can affect. Production capture, not a local screenshot, is the reviewer's source of truth.
  </Step>

  <Step title="Publish artifact and workplan to EyeParity">
    Publish the capture artifact and its workplan to EyeParity. Link the EyeParity artifact from the merge request.
  </Step>

  <Step title="Run vision analysis">
    Trigger EyeParity vision analysis on the artifact. Review every finding it produces.
  </Step>

  <Step title="Corroborate deterministically with DIPP">
    Run DIPP deterministic corroboration against the same artifact. Vision analysis without DIPP corroboration is not sufficient proof.
  </Step>

  <Step title="Fix and rerun every red or high finding">
    For every red or high-severity finding from vision analysis or DIPP, ship a fix and rerun the full capture, publish, analyze, and corroborate loop. Do not waive findings.
  </Step>
</Steps>

<Note>
  EyeParity's production route is the reviewer's source of truth. Screenshots pasted into merge requests, local previews, and dev captures do not substitute for the EyeParity artifact.
</Note>

## Related

<CardGroup cols={2}>
  <Card title="Frontend architecture" icon="cubes" href="/platform/frontend-architecture">
    Match each change to the package it lives in and the surfaces that need proof.
  </Card>

  <Card title="Routing and environments" icon="route" href="/platform/routing-and-environments">
    Confirm you captured the correct production origin for the affected product.
  </Card>
</CardGroup>