fix(cli): add V2 asset support for logs and check-run-data retrieval [AI-61]

[martzoukos]

Summary

• Add V2 asset handling to packages/cli/src/rest/assets.ts so that checkly test --verbose displays logs and check-run-data for accounts using the LEGACY_RUNNER_ASSETS_V2 feature flag
• Fix config.get crash in empty-account e2e tests by guarding with config.has

Problem

monorepo#596 (feat: add support for v2 assets for BROWSER/MULTI_STEP checks) changed how runners store check result assets. V1 uploaded each asset as a separate S3 file (logPath, checkRunDataPath, etc.). V2 bundles everything into a single ZIP archive with byte-range offsets per entry.

The runner side was updated but the CLI was not. The getAssets method in assets.ts only handled V1 and V3 formats — when it received a V2 asset, key stayed undefined and the method silently returned nothing. This meant npx checkly test --verbose showed no logs or check-run-data for BrowserCheck and MultiStepCheck runs on affected accounts.

The V2 types (CheckResultAssetsV2, isCheckResultAssetsV2) were already defined in the file but never wired into the retrieval logic.

Fix

Added a getAssetsV2 method that:

1. Looks up the requested entry by name in the entries array (logs.txt for logs, check-run-data.json for check-run-data)
2. Fetches a presigned S3 URL via the existing /redirect?link=true endpoint
3. Uses an HTTP Range request to fetch just the needed bytes from the ZIP
4. Parses and returns the JSON

[martzoukos]

Ok, let's wait for the tests to pass in the PR, first. 😅

[sorccu]

Those failures indicate an incompatible backend change. If accidental, this PR is not needed and the backend needs to be fixed. If on purpose, then we need to rethink these tests.

[martzoukos]

Those failures indicate an incompatible backend change. If accidental, this PR is not needed and the backend needs to be fixed. If on purpose, then we need to rethink these tests.

Fixing the source is definitely the way to go, if the tests broke suddenly and recently. Thank you for researching that.

[martzoukos]

Hello again @sorccu ,

armed with more context, I come back with new changes:

• I kept the packages/cli/e2e/__tests__/checks-empty-account.spec.ts changes for better handling of the case when there are no CHECKLY_EMPTY_API_KEY and CHECKLY_EMPTY_ACCOUNT_ID env vars. The code currently throws and this is reported as a failed test in CI. The reason why it might have not been picked up yet in the PRs is because these vars are present in Github Actions (but they might throw and cause some confusion in local runs).
• I replaced the previous test.spec.ts changes with the ones in packages/cli/src/rest/assets.ts. I'm obviously missing some context here but it seems that with a recent change on how we handle the fetching of the playwright tests' logs from S3 for accounts that are have the LEGACY_RUNNER_ASSETS_V2 flag enabled, we forgot to reflect this change in the CLI. With the help of claude, I made a addition for this, but I would appreciate a second set of eyes on this.

The tests seem to go green (save for the Windows one) with minimal changes to the cli tests code, so I think that might have a good solution on our hands.

Let me know what you think.

[MegaMaddin]

https://linear.app/checklyhq/issue/SIM-200/add-temporary-asset-handling-for-v2-assets

[sorccu]

Can you update the title and description to match the latest version of the PR?