test(e2e): make live image recognition non-blocking - #1363
Open
vercel-gh-bot-2[bot] wants to merge 1 commit into
Open
test(e2e): make live image recognition non-blocking#1363vercel-gh-bot-2[bot] wants to merge 1 commit into
vercel-gh-bot-2[bot] wants to merge 1 commit into
Conversation
Co-authored-by: ruiconti <1834568+ruiconti@users.noreply.github.com>
Contributor
|
The latest updates on your projects. Learn more about Vercel for GitHub.
1 Skipped Deployment
|
ruiconti
marked this pull request as ready for review
July 29, 2026 21:18
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The image content-parts eval currently turns live vision accuracy into a required PR gate. This keeps transport and tool execution checks blocking while recording color interpretation as a soft metric.
Motivation
The fixture renders a valid PNG and stores the exact stripe sequence in its action result. Live models still omit or misidentify a stripe. In one representative run, the image contained black, red, and orange while the model answered black and orange.
The observed direct mismatch rate was 22.2% locally and 23.9% on Vercel after the eval was introduced. Those failures measure model vision quality, not whether eve delivered the
toModelOutputcontent parts.Hypothesis
Because a soft assertion without a threshold stays visible in eval artifacts but does not fail
--strict, making the two color-recognition assertions soft should remove model-perception false negatives from the required gate. The eval still requires successful turns, no failed actions, one completedrender-stripescall with valid PNG output, and no second tool call during replay.Checks
pnpm exec oxfmt --check e2e/fixtures/agent-tools/evals/static-tools/to-model-output-content-parts.eval.tspnpm exec oxlint e2e/fixtures/agent-tools/evals/static-tools/to-model-output-content-parts.eval.tspnpm --filter eve exec vitest run --config vitest.unit.config.ts src/evals/runner/verdict.test.tspnpm --filter eve buildpnpm --filter agent-tools typecheckpnpm guard:invariantsThe fixture E2E itself remains CI-only, per the repository test policy.
Requested by: rui.lealsalvadorconti