Explanation: GitHub App

Hyaline can be installed via a GitHub App. The app has two primary flows: Checking pull requests to recommend documentation updates and extracting documentation when changes are merged.

Check PR

The check pull request flow, shown here in blue, starts with a developer opening a non-draft pull request against the default branch in a repository. The GitHub App is set up to receive this webhook event, and will dispatch a GitHub Action workflow to run a check on the changes made in that pull request. Note that this check uses the current documentation data set as an input, as it contains all of the documentation that has been extracted for the organization. Hyaline then posts a comment with the resulting recommendations back to the pull request for the developer to review. Each time the pull request is updated this cycle is triggered, and the pull request comment is updated with recommendations based on the latest changes.

Note that you can configure the GitHub App to ignore certain repositories if needed.

Extract

The extract workflow, shown here in green, starts with a developer merging a pull request into the default branch of a repository. The GitHub App is set up to receive this webhook event, and will dispatch a GitHub Action workflow to run an extract for the repository. This extracted documentation is then merged into the full set of existing documentation so that Check PR and other workflows have a data set containing the most recent extraction of the organization’s documentation.

Note that you can download the current documentation data set from the configuration repository for use with the MCP Server or for export.

The Hyaline GitHub App’s configuration is stored in a GitHub repository in your organization or personal account (see How To Install GitHub App for more information). This repository is where all of the workflows are run and artifacts are stored. This means that you have full access to and control of your data and the workflow actions run by Hyaline. The GitHub App comes with several built-in workflows, which include:

Doctor

The Doctor workflow checks your configuration and helps keep it up-to-date. It ensures the configuration is valid, opens a pull request for missing repository configurations, and keeps workflow inputs in sync as changes occur. You can run the doctor at any time, and the doctor will open a pull request in the configuration repository for you to review and merge if needed.

Extract All/Repo/Site

The Extract All, Extract Repo, and Extract Site workflows extract documentation using various configuration checked into the configuration repository.

The Extract All workflow will scan through the repos/ and sites/ directory and kick off an extract for each valid, non-disabled configuration it finds. It will (by default) trigger a merge workflow to merge the resulting extracted documentation together once the extractions are complete.

The Extract Repo workflow will extract a repository’s documentation based on the repo’s configuration file in the repos/ directory. It will (by default) trigger a merge workflow to merge the resulting extracted documentation together once the extractions are complete.

The Extract Site workflow will extract a website’s documentation based on the specified configuration file in the sites/ directory. It will (by default) trigger a merge workflow to merge the resulting extracted documentation together once the extractions are complete.

Run Audit

The Run Audit workflow runs an audit based on a configuration checked into the audits/ directory in the configuration repository. It will output the results of the audit as an artifact and as an annotation on the run itself.

Internal

Workflows marked as internal (starting with _) are used by the GitHub App to check pull requests, extract documentation, merge extracted documentation together, and perform audits. You should usually run one of the non-internal workflows instead of running an internal workflow directly.

Note that the artifact created by the _Merge workflow contains the merged current documentation data set, and can be downloaded and used with the MCP Server or exported into a variety of formats.