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

# Pipelines

> Turn failed CI pipeline runs into agent analysis and findings, and generate repository review reports

Review watches the CI pipelines — the automated build-and-test runs — of your connected repositories and turns each failed run into an agent-written analysis with findings, instead of leaving you to read raw job logs. This page covers turning pipeline analysis on, reviewing analyzed runs, and generating repository reports.

## Prerequisites

* A repository connected through [Review Setup](/guide/code-review/setup) with webhooks configured
* Permission to edit Review settings, for the toggles below
* A supported provider: GitHub, GitLab, Bitbucket, or Azure DevOps. AWS CodeCommit repositories do not support pipeline analysis.

## Turn on pipeline analysis

Pipeline analysis is a workspace-wide setting and is off by default.

<Steps>
  <Step title="Open Review settings">
    Go to **Review → Settings**.
  </Step>

  <Step title="Enable the toggle">
    Turn on the **Pipeline Analysis** switch. While it is off, the **Recent pipeline runs** section shows a banner with an **Enable Monitoring** button that does the same thing.

    **Success state:** new CI runs from your connected repositories appear under **Recent pipeline runs**, and failed runs are analyzed automatically.
  </Step>
</Steps>

## How a failed run becomes findings

1. Your provider sends a pipeline event, and Review records the run with its status and commit.
2. Only failed runs are eligible — successful, running, and pending runs are stored but never analyzed, and an already-analyzed run is not analyzed twice.
3. Review fetches the failed jobs and links the run to a reviewed merge request when one matches, so the analysis can use the review context of the change that broke it.
4. An agent analyzes the failed jobs in a dedicated conversation and writes root-cause findings, stored beside your normal review findings.

## Review a pipeline run

<Steps>
  <Step title="Open recent pipeline runs">
    Go to **Review → Settings** and scroll to **Recent pipeline runs**. Each row shows the repository, the pipeline number linked to your provider, the run status, and — for failed runs — the analysis status.
  </Step>

  <Step title="Start an analysis manually if needed">
    A failed run that has not been analyzed shows an **Analyze pipeline** button. Click it to start the analysis without waiting for the next event.
  </Step>

  <Step title="Open the pipeline detail">
    Open a run to see **Pipeline Jobs** — the stages and jobs with their statuses — and **Findings**, the agent's root-cause analysis for the failure.
  </Step>
</Steps>

The two status columns mean different things:

| Run status        | Meaning                                          |
| ----------------- | ------------------------------------------------ |
| Pending / Running | The pipeline has not finished on your provider   |
| Success           | The pipeline passed; no analysis happens         |
| Failed            | The pipeline failed and is eligible for analysis |
| Cancelled         | The pipeline was cancelled; no analysis happens  |

| Analysis status | Meaning                                                  |
| --------------- | -------------------------------------------------------- |
| Pending         | The failed run has not been analyzed yet                 |
| In Progress     | The analysis conversation is running                     |
| Completed       | The analysis finished and findings are available         |
| Failed          | Fetching the failed jobs or starting the analysis failed |

## Generate repository reports

Beyond per-run analysis, Review produces period-level repository reports: two PDFs per repository and review period — a technical report for engineers and a product report for leaders. They live in **Review → Insights** on the **Reports** tab.

<Steps>
  <Step title="Open the Reports tab">
    Go to **Review → Insights** and select **Reports**. Reports are grouped by period, with each row showing the repository, generation time, trend badges, action counts, status, and PDF links.
  </Step>

  <Step title="Choose scheduled or on-demand generation">
    Turn on **Enable scheduled repo reports** to let CloudThinker generate them on its own schedule for your active repositories, or click **Generate Now** for an immediate run.
  </Step>

  <Step title="Download the PDFs">
    On a completed row, open the **Technical PDF** or **Product PDF**.

    **Success state:** the row shows a completed status with both PDFs available.
  </Step>
</Steps>

<Note>
  Scheduled runs notify workspace admins when at least one report pair is ready. A manual **Generate Now** run sends no notification, because you are already on the page.
</Note>

Scheduled generation runs weekly per repository. The PDFs stay on the **Reports** tab — they are not emailed — and turning the schedule off keeps existing reports available.

### Weekly digest email

Separately from repo reports, CloudThinker emails the workspace owner a **Weekly Review digest** — a PDF summarizing the past week's review activity across the workspace. A week with too little review activity is skipped rather than sent empty. The digest appears as the **Weekly Review digest** email category under Review in [notification settings](/guide/notifications).

## Next steps

<CardGroup cols={2}>
  <Card title="Releases" icon="rocket" href="/guide/code-review/releases">
    Score release risk and record Approve or Hold decisions
  </Card>

  <Card title="Review Insights" icon="chart-mixed" href="/guide/code-review/analytics">
    Track review coverage, recurring findings, and trends
  </Card>

  <Card title="Review Setup" icon="gear" href="/guide/code-review/setup">
    Connect repositories and configure review modes
  </Card>

  <Card title="Mention Commands" icon="at" href="/guide/code-review/mention-commands">
    Interact with CloudThinker from MR/PR comments
  </Card>
</CardGroup>