Repository navigation
feat: record the capture order of screenshots - #360
Merged
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
jsfez
marked this pull request as draft
August 10, 2026 15:53
jsfez
marked this pull request as ready for review
August 18, 2026 13:10
This was referenced Aug 18, 2026
4 tasks done
A test walks a journey and screenshots it step by step, but none of that sequence reaches Argos: the CLI discovers snapshot files by globbing a directory, which yields them alphabetically. Each screenshot now carries its 0-based position within its test in `capture.index`. Counters are scoped to a test run, so a retry numbers its screenshots from 0 again, and one index is taken per call rather than per file, so viewport variants of the same screen share it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The `BrowserCommands` augmentation still described the previous arity, so passing the capture index over the RPC failed to typecheck. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jsfez
force-pushed
the
feat/capture-index
branch
from
September 2, 2026 12:22
9a63cec to
3b77458
Compare
4 tasks done
jsfez
added a commit
to argos-ci/argos
that referenced
this pull request
Sep 9, 2026
* feat(screenshots): accept the capture order in screenshot metadata The SDKs record where each screenshot falls within its test (`metadata.capture.index`, argos-ci/argos-javascript#360, shipped), but the schema does not declare the field — and unknown keys are stripped, so the index has been dropped on arrival ever since, silently. Keep it, and expose it through GraphQL, so the review can follow a test in the order it walked. The schema gains a test of its own: the failure mode here is an omission, which nothing else would have caught. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat(build): show the journey of the active diff in a drawer A test walks a journey and screenshots it step by step, but the review shows each screenshot on its own: nothing says which journey a change belongs to, or where in it the change sits. A drawer above the viewer lays that journey out as a strip of steps, in the order the SDK recorded, and jumps between them. It is built from every diff of the build rather than the filtered list, so the screens that did not change are in it too and the ones that did carry a mark — a change reads against the journey it happened in. ⇧← / ⇧→ walk the same steps from the keyboard, with or without the drawer open. A journey is the test that took the screenshot (its title path), a step is a logical screen (the backend `variantKey`, which already folds the viewport and browser variants of one screen together). Nothing to configure: it comes from the metadata the SDK already sends. A test that captures a single screen is regular visual testing, not a journey, so neither the drawer nor its toggle appears for it. The strip stays within the active diff's variant, so moving between steps keeps one viewport and the boxes can take their shape from it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(e2e): cover the journey drawer of the build review A build whose screenshots come from a test walking a checkout, with the capture order recorded. The names are chosen so neither the alphabet (cart, confirmation, payment, shipping) nor the diff score matches the journey (cart, shipping, payment, confirmation): a drawer that reads in order can only be reading the metadata. Two of the four steps are unchanged, so the journey is only whole if it is built from every diff rather than the reviewable ones. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(seeds): give the dev database a journey to review The drawer needs a test that captured more than one screen, and no dev scenario had one — every seeded test captures exactly one screenshot, so the feature was invisible locally. The checkout journey the e2e spec uses lands in the dev database too, on a project of its own: it owns build 1 and its buckets, which would collide with the builds `bigProject` already carries. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(build): stop clipping the descenders of journey step labels `truncate` hides overflow, so a line box the height of the font itself has nowhere to put a descender: "shipping" lost the tail of its g. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat(build): explain a journey step on hover Two things in the strip were unreadable on their own. The step label is cut to fit, so a long name ended in an ellipsis with no way to read it; hovering it now gives the whole variant key, which the label never shows in full since it only ever renders the last path segment. The dot had no legend anywhere. It now says what put it there, naming the statuses the way the rest of the review names them — "Needs review: Changed". It is 10px, too small to aim at, so the padding around it is the target and the offsets keep the dot itself where it was. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
A test walks a journey and screenshots it step by step, but none of that sequence reaches Argos: the CLI discovers snapshot files by globbing a directory, which yields them alphabetically. Each screenshot now carries its 0-based position within its test in
capture.index, which is what lets Argos replay a journey in the order it happened.Covered in Playwright, Cypress and Vitest — the three SDKs that report a test title path. Counters are scoped to a test run so a retry starts over at 0, and one index is taken per call rather than per file so viewport variants of the same screen share it.
🤖 Generated with Claude Code