Monorepo

Summary

DevLake's scope model is one scope per repository, so a repository that hosts several logically separate projects collapses into a single set of DORA numbers. The monorepo plugin splits such a repository into sub-projects and computes per-sub-project delivery metrics.

Like DORA and Linker, it is a metric plugin: it works on the domain layer after data collection, and runs after dora.

It attributes:

• Deployments to a sub-project by matching the name of the CI job that performed the deployment against a regular expression.
• Pull requests to a sub-project by matching the PR's labels.

From those two attributions it produces per-sub-project Deployment Frequency and Lead Time for Changes.

Change Failure Rate and Median Time to Restore Service are not computed per sub-project, because they require incident data and incidents are not attributed to sub-projects.

Supported Versions

The monorepo plugin was added after the v1.0.3-beta18 release and is available in all later releases.

Configuration

The plugin is configured per project, on the project's settings page.

1. Open Projects → your project → Settings.
2. Tick Enable Monorepo Sub-Projects.
3. Add one row per sub-project and fill in the three fields below.
4. Click Save, then re-run the project's blueprint.

Field	Description	Example
Sub-project name	Identifies the sub-project in the output tables and dashboards. Must be unique within the project.	serviceA
Labels	Comma-separated pull request labels that mark a PR as belonging to this sub-project. Matching is exact and case-sensitive.	serviceA,svc-a
Deploy job pattern	A Go regular expression matched against cicd_tasks.name to recognise this sub-project's deploy jobs.	^deploy-serviceA$

unattributed and All are reserved and cannot be used as sub-project names.

The same options can be set through the API or the Config UI's advanced mode as the project's monorepo metric plugin option:

{
  "pluginName": "monorepo",
  "enable": true,
  "pluginOption": {
    "subProjects": [
      {
        "name": "serviceA",
        "prLabels": ["serviceA", "svc-a"],
        "deployJobPattern": "^deploy-serviceA$"
      },
      {
        "name": "serviceB",
        "prLabels": ["serviceB"],
        "deployJobPattern": "^deploy-serviceB$"
      }
    ],
    "includeUnattributed": true
  }
}

includeUnattributed (default true) controls what happens to pull requests and deployments that belong to the project but match none of the configured sub-projects: with true they are recorded under the sentinel sub-project unattributed, so you can see how much of the repository is not yet classified; with false they are left unclassified.

How attribution works

Deployments

Every deployment of the project is joined to the CI/CD tasks of the pipeline that produced it, and each task of type Deployment has its name matched against each sub-project's deployJobPattern.

• A pipeline that runs the deploy jobs of several sub-projects produces one row per sub-project, so dashboards can GROUP BY sub_project without double counting.
• Deployments imported directly from a provider's deployment API (rather than generated from a pipeline) carry no job name and cannot be attributed.

Pull requests

All pull requests of the project are attributed — open, closed and merged — by label. A pull request belongs to exactly one sub-project: when it carries the labels of more than one, the first matching sub-project in the configured order wins, so order the list from most to least specific.

PR coding, pickup, review, deploy and cycle time are not recomputed by this plugin; they are taken from DORA's own project_pr_metrics rows and tagged with the sub-project.

Mismatch warnings

After tagging, the plugin cross-checks each PR's label-derived sub-project against the sub-project of the deployment that DORA says actually shipped it. Disagreements are logged as warnings and never fail the pipeline — labels always win. They usually point at a prLabels or deployJobPattern entry that needs fixing, though a pipeline deploying several sub-projects at once will produce such warnings legitimately.

Output

Table	Description
cicd_deployment_subprojects	Maps each deployment to the sub-project(s) it deployed.
pull_requests.sub_project	The sub-project a pull request belongs to.
pull_request_commits.sub_project	Mirrors the owning pull request's sub-project.
project_pr_metrics.sub_project	Tags DORA's per-PR metrics with the sub-project.

A NULL sub_project means the row was never classified — single-repo projects, or rows the plugin has not processed. Dashboards render these as the single implicit group All via COALESCE(sub_project, 'All').

Attribution is rebuilt from scratch on every run, so changing the configuration and re-running the blueprint reclassifies existing data.

备注

The plugin also populates monorepo_subproject_deployments and monorepo_subproject_pr_metrics. Both are deprecated and will be removed in a future release — query the tables above instead.

Dashboard

A built-in Grafana dashboard, Monorepo Sub-Projects, visualises the output:

• Deployment Count by Sub-Project
• Deployment Frequency over Time
• Change Lead Time Breakdown by Sub-Project
• Unattributed PRs
• Awaiting Deployment
• Attributed Deployments