GitHub Insights

Who is looking at a repository, where they came from, and what they opened — Traffic API. The same question a website asks of its access log, asked of a repository, and answered into an email instead of a browser tab.

Part of the GitHub family: github (issues and pull requests), github-publish (files), github-release (releases and assets), github-repo (repository administration).

Use case

You published a repository and announced it somewhere. Did anyone arrive? From which link? Did they read past the README? operation=summary returns the whole Traffic page as text — schedule it weekly and the answer arrives without you going to look.

Setup

1. Create a personal access token at https://github.com/settings/tokens with Administration: read and Contents: read on the repository. A classic token needs repo.
2. Store it once: Credentials → New → API Key, name it (e.g. github-blueisle), paste the token.
3. Name that credential on every github-insights task.

Operations

Two limits that are not bugs

Traffic needs push access. A token that can read a public repository still gets 403 here — these numbers belong to the owner, not the public. The tool says exactly that when it happens.

The window is fourteen days, fixed. There is no date-range parameter and GitHub retains nothing older. A longer trend has to be accumulated: run summary on a weekly schedule and keep what it publishes. That is the difference between having a launch-week number and having a launch-week number to compare against.

Pipeline

1. run (python, github_insights.py) — resolves the token from the bound credential, calls the traffic endpoints (four of them for summary), renders the report, publishes the counts.

Publishes

• result — the report text (summary) or the raw JSON body
• status — HTTP status code as a string
• views, uniqueViews, clones, uniqueCloners — the 14-day totals as numbers, so a workflow can compare them against last week's without parsing the report

Customization

• per: "week" groups views and clones by week instead of day — steadier for a monthly digest.
• Pair with the message tool to mail the report, exactly as the website traffic workflows do.
• Counts include automated fetches. GitHub filters some bots and not all, so treat the unique figures as the honest ones.

Source

tool.iml (definition) · github_insights.py (implementation) · manifest.yaml (catalog entry)

Publishes

The keys a later task may reference as {{taskName.key}}. Referencing anything not listed here resolves to nothing at run time.

• clones
• result
• status
• uniqueCloners
• uniqueViews
• views

Input Properties

Every property this tool accepts, from its own tool.iml.