Importing historical test data

Backfill a history repository with test results from previous workflow runs on a source repository. The import tool downloads raw test-result artifacts (TRX, JUnit, NUnit, xUnit), parses them with the same parsers as the action, converts them to history JSON, and opens a pull request.

Prerequisites

• gh and git on your PATH
• Node.js 20+ (for the TypeScript import engine)
• Authenticated gh auth login (or GITHUB_TOKEN)
• Read access to the source repository's Actions runs and artifacts
• Write access to the history repository (for the PR branch)
• Workflows that upload test result files as artifacts (for example test-results containing *.trx or *.xml)

GitHub Actions artifact retention applies: runs whose artifacts have expired cannot be imported.

Quick start

From an actions-insights checkout:

bash scripts/import-history-repo.sh import owner/my-app owner/my-history \
  --artifact-name test-results --limit 20

Or via the optional GitHub CLI extension:

gh actions-insights import owner/my-app owner/my-history --artifact-name test-results

From the monorepo directly (after cloning the history repo yourself):

npm run import-history -- owner/my-app --repo-path /path/to/history-repo \
  --artifact-name test-results

Confirmation and dry-run

Like init and update, the bash wrapper shows a summary and asks Proceed? [Y]es / [n]o / [e]dit: before cloning or writing. Use --dry-run to preview without changes, or --yes for CI.

Flags

Flag	Default	Description
--artifact-name <name>	auto-detect	Exact artifact name (repeatable)
--artifact-pattern <glob>	—	Artifact name glob (repeatable)
--test-results-glob	**/*.{trx,xml}	Glob inside downloaded artifact
--workflow <name|id>	all	Filter workflow runs
--branch <name>	all	Filter by branch
--since <ISO date>	none	Only runs after date
--limit <n>	50	Max runs to scan
--history-limit	20	Branch history limit
--retain-days	30	Retention window
--data-path	data	History data root

When no artifact name or pattern is given, the importer tries each artifact on a run and uses the first one that contains parseable test files.

Workflow example

Upload test results as artifacts so historical runs can be imported later:

- run: dotnet test --logger "trx;LogFileName=results.trx"
- uses: actions/upload-artifact@v4
  if: always()
  with:
    name: test-results
    path: '**/*.trx'

What gets written

For each imported run, the tool writes the same JSON files as a live history-enabled publish:

• data/repositories.json
• data/repositories/{owner.repo}/metadata.json, branches.json, tests.json
• data/repositories/{owner.repo}/branches/{branchKey}/latest.json, history.json
• data/repositories/{owner.repo}/branches/{branchKey}/runs/{timestamp-runId}.json

Runs already present in the history repository (matched by workflowRunId) are skipped.

Imported JSON uses schema v2 (minified, class dictionaries, name index). See Data model.

Related

• Migration guide — moving from cache-only history
• Deployment — initializing and updating the history repository