test: claim-review MVP end-to-end integration coverage

[enlorik]

Summary

• Adds test/search_screen_claim_review_e2e_test.dart — 2 focused widget tests, no production code changed.
• Exercises the full configured MVP flow through SearchScreen using fakes only (no network, no real API keys).

Test 1 — full MVP flow:

• Ask question → provider called once, local evidence retrieved → groups render with provider label → no setup-required card
• Default selection correct (new + betterSource selected, alreadyKnown not)
• Toggle better-source off → generate draft A (new claim only) → verify content, exclusions, citation title/URL, no raw prose, no provider name
• Save draft A → verify exactly 1 insert, source URL preserved, no secrets in content
• Double-persistence guard: append after save is blocked; repeated save returns ignored, no second insert
• Toggle better-source back on → draft cleared → generate draft B (new + better-source) → verify differs from draft A
• Append draft B to target note → 1 update, original content preserved, draft B appended, embeddingPending set, no duplicate on repeat
• Discard → session cleared (idle, no groups, no draft, no label), repository records intact

Test 2 — alreadyKnown tampering regression:

• Review with only an alreadyKnown claim
• Directly call notifier.toggle(knownClaimId) to simulate corrupted selection
• selectedSaveableItems remains empty (canBeSaved = false)
• generateDraft() produces null or non-saveable draft
• saveAsNewNote() returns ignored, no insert
• appendToExistingNote() is a no-op, no update

Stale #70 assumptions fixed

Old assumption	Why incompatible	Fix
_FixedGroundedAnswerProvider missing isConfigured	isConfigured added to abstract class in #73	Added bool get isConfigured => true
Same draft saved then appended	isDraftAlreadySaved guard prevents this (correct behavior)	Append path now uses distinct draft B

Verification

• flutter analyze → No issues found
• flutter test test/search_screen_claim_review_e2e_test.dart --reporter expanded → 2/2 pass
• flutter test --reporter expanded → 589/589 pass
• 1 commit, 1 file changed (+533 lines)

@codex review

🤖 Generated with Claude Code

[enlorik]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Nice work!

Reviewed commit: b21a1b71c7

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".