History Pages: Test Convergence Plan

Date: 2026-02-28 (updated) Context: Notebooks→Pages merge complete. UI convergence complete. Cleanup rename (notebook→page everywhere) complete. Two page types: regular (slug-based, publishable) and history-attached (history_id, revisions with edit_source, agent chat).

Current State

Naming Cleanup: Done

All code uses “page” terminology. Specifically:

• Backend: PageAssistantAgent, AgentType.PAGE_ASSISTANT, create_page_chat(), page_id FK
• Frontend: PageEditor/ directory, usePageEditorStore, pages.ts API client, PageChatPanel.vue
• Selenium: navigates_galaxy.py helpers all say “page”, navigation.yml selectors renamed
• Unit tests: test_chat_manager.py uses page_id, test_agents.py uses PageAssistantAgent
• No stale “notebook” references remain (except unrelated Jupyter body.notebook_app)

API Tests

Selenium Tests

Frontend Vitest

Comprehensive (~2,650+ lines, 150+ tests). All use page terminology.

Backend Unit Tests

Gaps: API Tests

Gap A1: History-Attached Page CRUD (HIGH)

Zero API integration tests for history-page workflow. Populators ready, just need tests.

Gap A2: edit_source Tracking (HIGH)

No API tests verify edit_source is recorded on revisions.

Gap A3: Revision Endpoints (MEDIUM)

test_page_revisions.py tests create + list + permissions. Missing:

Gap A4: History-Page Permissions (MEDIUM)

Regular page permission tests exist (403 on unowned). History pages need:

Gap A5: Cross-Type Validation (LOW)

Gaps: Selenium Tests

Gap S1: Regular Pages Are Undertested (MEDIUM)

Only 3 selenium tests for regular pages vs 24 for history-attached. Missing:

• Page editing (update existing content, re-save)
• Page deletion via grid UI confirmation flow
• Published page viewing by another user

Not critical — API tests cover the backend well. But post-convergence, the unified PageEditorView is used for both types, so regular-page editing through the new editor is untested at the E2E level.

Gap S2: Standalone Page Uses Unified Editor (MEDIUM — new gap from UI convergence)

The UI convergence replaced PageEditorMarkdown.vue with PageEditorView. No selenium test verifies that editing a regular page through the new unified editor works. This is a new code path — the old PageEditorMarkdown had its own save logic (POST /api/pages/{id}/revisions), the new one uses the store’s savePage().

Suggested test: test_edit_standalone_page_via_unified_editor — create page, navigate to /pages/:id/edit, modify content, save, verify persistence.

Gap S3: Chat Panel E2E (LOW)

No selenium tests for the chat panel. Vitest suite (PageChatPanel.test.ts, 327 lines) covers message flow, persistence, and error handling well. Selenium would require a running LLM backend — impractical for CI. Defer to env-gated integration tests.

Gap S4: Cross-Type UI Behavior (LOW)

No selenium tests verify:

• Regular pages show Permissions button but not history-specific controls
• History pages show chat/revisions but not slug/publish controls

Mostly a UI convergence correctness concern. Vitest (PageEditorView.test.ts) covers toolbar differences by mode.

Recommendations

Priority 1: API Tests for History Pages

Add TestHistoryPagesApi class — either in test_pages.py or new test_history_pages_api.py. Cover gaps A1 + A2. This is the single biggest hole: the entire history-page API path has zero integration test coverage.

~15 new test methods, ~250 lines. Populators ready — no infrastructure work needed.

Priority 2: Revision API Completion

Extend test_page_revisions.py with gap A3 (get single revision, revert, revert behavior). These endpoints are exercised by the frontend but have no direct API test coverage.

~4 new test methods, ~80 lines.

Priority 3: History-Page Permissions

Add gap A4 tests. Shared-history read access is the most important — it’s how collaborators see each other’s pages.

~3 new test methods, ~60 lines.

Priority 4: Standalone Page Selenium Test

One selenium test verifying the unified editor works for regular pages (gap S2). This catches regressions from the PageEditorMarkdown → PageEditorView replacement.

~1 test method, ~30 lines.

Priority 5: Cross-Type Validation

Gap A5 tests. Low-risk but good for documenting the contract between the two page types.

~3 test methods, ~40 lines.

Test Infrastructure

Populators support all needed operations — no new infrastructure required:

# History page CRUD
new_history_page(history_id, title, content, content_format)
get_history_page(page_id)
list_history_pages(history_id)
update_history_page(page_id, content, title, edit_source)
delete_history_page(page_id)

# Revisions
list_page_revisions(page_id)
revert_page_revision(page_id, revision_id)

# Sharing
make_page_public(page_id)

Selenium navigation helpers:

navigate_to_history_pages()
history_page_create()
history_page_editor_set_content(content)
history_page_save()
history_page_manage()
history_page_open_revisions()
history_page_assert_revision_count(n)
history_page_rename(new_name)
# ... 11 total

Unresolved Questions

1. History-page API tests — test_pages.py (add class) or separate test_history_pages_api.py?
2. Standalone page selenium — add to existing test_pages.py or new file?
3. Export-to-Page flow (history-page → published page with slug) — test now or defer as unimplemented feature?