[hyoban]

Summary

Continues #44243, #44260, and #44274. Two workflow canvas suites mocked the hooks that own the behavior they claimed to test, so they kept passing even when that behavior broke. They now run against the real workflow and ReactFlow stores.

• workflow/operator/__tests__/control.spec.tsx: drops the mocks of useWorkflowMoveMode, useNodesReadOnly, useWorkflowOrganize, and useOperator. The tests now check what the user sees and what lands in the store: the control mode switches between pointer, hand, and comment, adding a note creates a candidate credited to the current user, organizing starts the layout animation, read-only mode leaves the canvas unchanged, comment mode is blocked during a run, and the comment button is hidden when collaboration is off. The add-block and export-image children stay mocked because they have their own suites and need their own service setup.
• workflow/note-node/__tests__/index.spec.tsx: removes six mocks of useNote and useNodesInteractions that no test ever asserted against, plus the collaboration-manager mock and an as unknown as cast. Adds behavior coverage through the real note menu: hiding the author persists the preference, and delete removes the note from the canvas.

Further batches will be pushed to this branch.

Batch 2: draft sync and note hooks on real stores

• The use-nodes-sync-draft suites for workflow apps, knowledge pipelines, and snippets replaced the ReactFlow store, workflow store, features store, and useNodesReadOnly with hand-written stand-ins and asserted on their setters. They now run on real stores through a new renderWorkflowCanvasHook helper in workflow-test-env. The helper seeds a real ReactFlow store without mounting a canvas and exposes setCanvas() for teardown and queueing scenarios. Only the network client, keepalive transport, collaboration session, and draft refresh stay mocked.
• The assertions now check what the store actually holds. For example, draftUpdatedAt is stored in milliseconds. The old mocks hid this, and read-only mode was never exercised before. Repeated cases (error-code refresh, page-close flush matrix, read-only/empty/temporary-only) are merged into it.each tables. Fixtures use createNode/createEdge instead of casts.
• note-node/__tests__/hooks.spec.tsx no longer mocks node-data update and workflow history. It checks the real note data and the history entry. The show-author path is covered through the note menu in the batch 1 suite.

Batch 3: pipeline and workflow run hooks on real stores

• The rag-pipeline hooks use-pipeline-refresh-draft, -init, -start-run, -run, and use-DSL, plus the workflow-app hooks use-workflow-init and use-workflow-start-run, no longer stub the workflow store, the ReactFlow store, the features store, useWorkflowInteractions, or useInputFieldPanel. They now assert what the real stores hold: syncing flags, the draft hash, masked secrets, pipeline variables, panel visibility, canvas selection, and run state. Draft sync, the run stream, canvas update, and the run-event reducers stay mocked because each has its own suite.
• use-pipeline-run: thirteen tests that each checked one callback being passed through are merged into one it.each table. Abort behavior is now tested with a real AbortController, both for stop and for starting a new run.
• use-pipeline-init uses the real DatasetDetailContext and the real knowledge-base template instead of mocking them.
• use-update-dsl-modal.ts called useWorkflowStore() but never used the value. The only reason it appeared in a useCallback dependency list was that pipelineId was missing from it. The unused store call is removed and pipelineId is added to the dependency list. The test's store mock is removed as well.
• use-file-settings.spec.ts only checked that a constant came back unchanged and is deleted.

Batch 4: snippet run and pipeline panels on real stores

• snippets/hooks/use-snippet-run.spec.ts replaced its stand-ins for the ReactFlow store and viewport with a real canvas, and now covers more of the hook. Previously it only checked stop. It now also covers draft backup and restore, restoring a published snippet, the run lifecycle (one it.each table for the forwarded stream callbacks), run-history and inspect refreshes, and abort on stop or on a new run.
• The rag-pipeline component suites publish-toast, the input-field useFloatingRight, the test-run header, test-run result, and header run-mode now run on real workflow, hooks, and ReactFlow stores. Run-mode passes its start and stop handlers through hooksStoreProps instead of module mocks and uses the real event emitter. The new tests cover the case where the user has no run access and the case where another panel requests a stop. update-dsl-modal.spec drops a dead store mock.
• I left workflow-app/hooks/use-workflow-run.spec.tsx unchanged because it already mocks only network, analytics, and audio boundaries.
• Source fixes for stale closures in useCallback dependency lists:
  • use-get-run-and-trace-url.ts: removed the unused useWorkflowStore() call. pipelineId was missing from the dependency list.
  • use-pipeline-run.ts handleStopRun: depended on [workflowStore] but reads pipelineId.
  • test-run/header.tsx: handleCancelDebugAndPreviewPanel was missing from the dependency list.
  • Separate commit fix(web): include editor resource ids…: when pipelineId and appId moved from the workflow store to the editor-resource atom, several callbacks kept their old dependency lists. They could then call stale /rag/pipelines/<old> or /apps/<old> URLs if the resource changed without a remount. The commit adds the ids to the useCallback dependency lists in rag-pipeline use-DSL, use-pipeline-run (handleRun), use-pipeline-refresh-draft, and use-nodes-sync-draft, and in workflow-app use-workflow-refresh-draft and use-nodes-sync-draft (three callbacks). It also adds closeAllInputFieldPanels to use-pipeline-start-run and removes two unused store dependencies. No behavior changes in the current single-mount flow.

Batch 5: hooks-store handlers, real event emitter, remaining store stand-ins

• Snippet run-mode and snippet-run-panel now get their run and stop handlers through hooksStoreProps instead of module-mocking use-workflow-run / use-workflow-start-run. The run panel's close button now goes through the real useWorkflowInteractions and the test checks that the panel actually closes.
• chat-wrapper.spec used to fake eventEmitter.useSubscription and call the captured callback directly. It now uses the real EventEmitterContextProvider, emits EVENT_WORKFLOW_STOP, and checks that the stop happens. layout-main.spec drops an emitter mock that only repeated the context default.
• help-line and variable-assigner/hooks no longer stub the ReactFlow store, the workflow store, or useNodeDataUpdate. Assertions now check real node data and store state. New cases cover removing the guides once alignment ends and the hover-leave case when no connection is in progress.
• Test-only batch; no source changes.

Validation

• vp test run --project unit app/components/workflow/operator app/components/workflow/note-node: 18 files / 77 tests passed, three runs in a row.
• vp check on the changed files and the vp staged commit hook passed.
• Batch 2: the touched suites (22 files / 135 tests) and the vp staged commit hook passed locally. CI is the full check.
• Batch 3: the touched suites (16 files / 135 tests) and the vp staged commit hook passed locally.
• Batch 4: rag-pipeline and snippets hook suites (62 files / 343 tests) and the vp staged commit hook passed locally. CI on batches 1–3 is green.
• Batch 5: touched suites (31 files / 176 tests) and the vp staged commit hook passed locally.

Batch 6

• block-selector/agent-selector: drop a dead useHooksStore mock (component never reads the hooks store).
• header/run-mode, panel/version-history-panel: replace useWorkflowRun/useWorkflowStartRun/useDSL/useWorkflowRefreshDraft module mocks with handlers passed through hooksStoreProps.
• node-actions-menu: select/copy/duplicate/delete now run against a real canvas and block catalog, asserting canvas nodes and clipboard instead of hook calls; the duplicate interaction test in details.spec is replaced by a help-link check.
• block-selector/tabs reviewed and kept as is (only network mocked).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 88.60%. Comparing base (aa0924a) to head (965c055).

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #44287      +/-   ##
==========================================
+ Coverage   88.57%   88.60%   +0.03%     
==========================================
  Files        5507     5507              
  Lines      328615   328613       -2     
  Branches    65304    65304              
==========================================
+ Hits       291063   291164     +101     
+ Misses      32359    32256     -103     
  Partials     5193     5193              

Flag	Coverage Δ
dify-ui	91.67% <ø> (ø)
web	88.39% <ø> (+0.07%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.