[mmabrouk]

Context

In the /m New Auto Evaluation drawer, picking an application after you type a name cleared the name. Release QA for v0.128.9 found this. The drawer cleared the name because picking an app rebuilt the whole draft from empty, and the name went with it.

Changes

A name that you type now stays when you pick an app, pick a different app or remove the app. The name auto-fills only while you have not typed one:

• Typed "my capitals check", then picked Capital finder and a revision. Before: the name field was empty. After: it still reads "my capitals check".
• A name the drawer made itself still gives way when you pick another app, and it fills again from the new picks.
• If you clear the field, the auto name comes back when you leave the field.

Note on /w: in the old app, handleAppSelection in NewEvaluationModalInner.tsx also calls setEvaluationName("") when the app changes. This PR follows the rule the release QA asked for, so /m now keeps a typed name.

The drawer is only reachable with beta developer mode on (evaluations is a hidden developer view). Nothing changes for /m without developer mode.

Tests / notes

• New tests in newEvaluationDraft.test.ts: "A typed name stays when an app is picked, changed or cleared, and later picks keep it" fails before the fix. "An auto name gives way to the next app, and a cleared name auto-fills again" covers the auto-fill.
• newEvaluationDraft.test.ts and newEvaluationSheet.render.test.tsx pass (42 tests). /m types pass, eslint passes, and prettier and gitleaks ran in the commit hook.
• Checked live on a parity stack: typed a name, picked Capital finder and a revision, and the name stayed.

What to QA

• Evaluations > New evaluation > Auto evaluation. Type a name, pick an app, then a revision and a test set. The name you typed stays.
• Open a fresh drawer and pick an app, a revision and a test set without typing. The name fills in by itself.
• Clear the name field and click outside it. The auto name comes back.

[coderabbitai]

No actionable comments were generated in the recent review.

Summary

Summary by CodeRabbit

• Bug Fixes
  • Manually entered evaluation names are preserved when changing or clearing the selected app. Automatically generated names continue to update with app selections.
  • Clearing an app returns the draft to its default section; selecting an app opens the revision section.

The evaluation draft reducer now preserves a manually typed name when an app is selected, changed, or cleared. Tests cover name preservation and updates to auto-generated names.

Changes: Evaluation draft naming

Layer / File(s)	Summary
Preserve typed names across draft changes web/mobile/src/features/evaluations/new/newEvaluationDraft.ts, web/mobile/tests/unit/newEvaluationDraft.test.ts	The reducer preserves a manually typed name when an app is changed or cleared. Tests cover typed names across subsequent picks and the behavior of auto-generated names.

Estimated code review effort: 2 (Simple) | ~8 minutes

Possibly related PRs

• Agenta-AI/agenta#8149: Adds draft input detection and a discard confirmation that reads the draft name.

Merge Risk: Minimal · up to 4cd8e

Typed names are preserved in the common app-selection workflow. No merge-blocking risk remains.

Pre-merge checks: 5

Check name	Status	Explanation
Title check		The title clearly describes the main change: preserving a typed evaluation name when an app is picked.
Description check		The description matches the changes to typed and automatically generated evaluation names and their tests.
Docstring Coverage		No eligible functions found in the changed files. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentat…
Linked Issues check		Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check		Check skipped because no linked issues were found for this pull request.

Full details: Docstring Coverage

Explanation

No eligible functions found in the changed files. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentation-only edits and deletions within surviving functions are included; unchanged diff context and fully deleted functions are excluded. Analyzed 0 functions across 2 files.

Finishing Touches

Generate docstrings

• Commit to this branch
• Create a new PR

Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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

Recent review info

Run configuration

• Source mode: Auto
• Repository: Agenta-AI/agenta
• Branch: fix/m-new-evaluation-keep-typed-name
• File: .coderabbit.yaml
• Commit: 4cd8e37
• Layers: Repository YAML (base), Organization UI (inherited)
• Review profile: CHILL
• Plan: Advanced
• Run ID: 7c80853b-4e0f-4e92-b12d-d80f94f141bf

Commits

Reviewing files that changed from the base of the PR and between 53942d0 and 4cd8e37.

Files selected for processing: 2

• web/mobile/src/features/evaluations/new/newEvaluationDraft.ts
• web/mobile/tests/unit/newEvaluationDraft.test.ts

Included review availability: This review used your included allowance. Your plan provides up to 8 included reviews per hour; 3 remain after this review.

[github-actions]

Railway Preview Environment

Item	Value
Status	stopped
Commit	4cd8e37de04e8cffd929c8e3bf8051c1587a23de
Manual expiry (UTC)	Not granted
CI deadline (UTC)	Not granted
Workflow	https://github.com/Agenta-AI/agenta/actions/runs/42258538320

Dependent tests finished. Post /preview for manual review. Environment absence verified; retained storage requires separate provider accounting.

Post /preview as a new comment to request a one-hour preview. Repeating it while active does not extend the hour.
Preview data is disposable. Expiry is enforced by the next successful cleanup sweep, not an exact timer.
Environment deletion stops its compute. Provider-retained storage and permanent services may still be billed.

[mmabrouk]

🤖 The AI agent says: @coderabbitai review

[coderabbitai]

@mmabrouk I’ll start a review of the PR.

Action not completed

Already reviewed the last commit. Use @coderabbitai full review to rerun a review of the entire changeset.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.