refactor(frontend): drop per-project fan-out from all batch fetchers

[mmabrouk]

Context

While reviewing #4634, @junaway flagged the artifact fetcher's shape: it grouped requests into a Map keyed by project_id and ran one query per project inside a Promise.all. That structure handles a state that cannot exist. The web app holds exactly one project_id in scope at any time, and the /query endpoints accept multiple ids, so a batch of coalesced requests always resolves with a single call.

The interesting part is where that shape came from. It was not invented for #4634; it was copied from the fetchers sitting right next to it. A grep for the pattern found it in 13 batch fetchers across 9 files: workflow revisions, latest revisions, traces, spans, testsets, testset revisions (twice), testcases (twice), evaluation runs, annotations, and environments. Every one of them defends against a multi-project batch that no caller can produce. Each copy also drags in real complexity: nested grouping maps, per-group error handling, and in one case a per-project queryClient set that exists only to serve the impossible case.

This is how a wrong pattern propagates. "Match the surrounding code" is the default instinct for any contributor (or agent) writing a new fetcher, and the surrounding code taught the wrong thing. #4634 fixed its own fetcher after review; this PR removes the remaining 12 copies and writes the invariant down so the next fetcher starts from the right shape.

Changes

Every batchFn now follows one shape:

1. Pre-fill the results map with the miss value, so error paths need no per-key cleanup.
2. Take project_id from the first request. If any other request disagrees, throw. A multi-project batch means a bug upstream, and per @junaway's review it must fail rather than be silently accommodated.
3. Resolve all ids with a single query.

Before (the shape every fetcher had):

const byProject = new Map<string, {ids: string[]; keys: string[]}>()
requests.forEach((req) => { /* bucket into byProject */ })
await Promise.all(
    Array.from(byProject.entries()).map(async ([projectId, group]) => {
        /* one query per project, per-group error handling */
    }),
)

After:

const projectId = requests[0]?.projectId
if (!projectId) return results
if (requests.some((req) => req.projectId !== projectId)) {
    throw new Error("…: requests span multiple projects")
}
/* one query with all ids */

Behavior that was per-fetcher stays: cache short-circuits, fallback-to-individual-fetch paths, and cache priming are preserved, just no longer duplicated per project group. Grouping by other dimensions survives where it is real: the testcase entity fetcher still groups its cache lookup by revisionId, because the paginated cache is genuinely revision-scoped. The rule is only that the project dimension is singular.

web/AGENTS.md now states the invariant under Data fetching ("Single project scope"), so the convention no longer lives only in the code it was just removed from.

Tests

• tsc --noEmit clean on agenta-entities; zero errors in the two touched web/oss files (the workspace has unrelated pre-existing errors elsewhere).
• All 656 agenta-entities unit tests pass.
• eslint clean on all nine touched source files.
• Net −186 lines.

What to QA

This is a refactor with no intended behavior change, so the check is regression-only: the touched fetchers feed most entity displays. Browse the evaluators table, a testset and its revisions, an evaluation run detail, the annotation queue, a trace drawer, and the deployments page. Everything should load exactly as before, with no new errors in the console.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
agenta-documentation	Ready	Preview, Comment	Jun 11, 2026 10:22am

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: accd39a4-59e2-40b9-81f1-2727e2eb5bad

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch refactor/single-project-batch-fetchers

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}