fix(flow-functionality): enter the flow through openFlowById (#1342)

[Victor-w-Madeira]

Fixes #1342. Closes #1342.

Problem

#1336's fix re-opened the flow under test with a local page.goto('/flow/{id}') plus a canvas wait. It works and it is validated — but it is the fourth copy of the block helpers/flows/open-flow-by-id.ts (#1214) was extracted to stop, added by the very PR that was fixing a parallel-safety bug of the kind that helper documents. Self-reported debt, not a review catch.

Two guarantees the copy did not have, both live for this spec:

• The onboarding overlay. openFlowById seeds the discovery flag before the document load. Upstream arms the assistant tooltip at canvas mount + 10 s and paints it over the canvas-controls bar — which is exactly what adjustScreenView clicks (Measure whether the two remaining dismissOnboardingIfPresent probes actually catch the tooltip #1220's measurement, not an inference). This test calls adjustScreenView four times across a ~15–25 s run, and the CI artifact that opened [Daily #1330] auto-save-off — reopening the just-created flow's card intermittently times out at 45 s (recurrent 2×) #1336 shows that tooltip painted over the canvas.
• The writability gate. openFlowById refuses to hand back an editor while POST /api/v1/authz/me/permissions is still in flight — the renameFlow residual flake: read-only save button and missing flow_name header (~7%, not the #995 clobber) #1005 window in which a mutation is silently swallowed. This spec adds a component immediately after two of the three re-opens.

Fix

Both re-opens now enter through openFlowById. Two deliberate details:

• The seed is called at the top of the test, not left to the helper's own call. addInitScript applies only to loads that follow it, and the first editing phase — two adjustScreenView calls plus the on-canvas save click — happens before any re-open, inside the mount + 10 s window.
• The wait on the flow's own GET /api/v1/flows/{id} stays. This is the part the issue flagged as needing care and it is load-bearing: openFlowById returns on canvas_controls_dropdown + writability, and neither implies the graph has been applied. The discard assertion (div-generic-node count === 0) is the one check that passes vacuously on a canvas that has not painted its nodes yet, so the count is ordered after that GET. The response promise is registered before the helper's navigation and awaited after it.

Scope note

No assertion changed; no change to the exit path (expectLeftEditor is untouched — leaveFlowEditor is not applicable here, it is built for autosave mode where the blocker dialog renders no buttons, whereas this spec must deliberately click "Exit Anyway" / "Save And Exit", which is the behaviour under test).

Validation (nightly 1.12.0.dev18, --retries=0)

• npm run typecheck ✅ · npx eslint <spec> ✅ (0 errors)
• 3/3 and 2/2 clean serial runs; zero 🚨 Backend Error ✅
• 8/8 under --workers=4 --repeat-each=8 — the burst that produced 3/8 failures before [Daily #1330] auto-save-off — reopening the just-created flow's card intermittently times out at 45 s (recurrent 2×) #1336's fix ✅
• No leak: 27 flows before and after, 0 New Flow orphans ✅
• Force-fails, executed:
  • expect(chatInputNode).toBe(0) → toBe(1) ⇒ failed (Expected: 1, Received: 0) — the discard assertion the issue asked to re-prove is still falsifiable through the new entry
  • expect(nodeCount).toBe(2) → toBe(3) ⇒ failed (Expected: 3, Received: 2)
  • identity: every re-open pointed at the bootstrap placeholder id (deleted by Langflow) ⇒ failed in both layers — the flow-GET gate at 45 s and the helper's own canvas_controls_dropdown assertion (element(s) not found), with the 404 logged
  • mutations reverted (grep → 0 hits) + final green run ✅

Stated rather than claimed: the placement of seedAssistantDiscovered at the top of the test is not independently falsifiable by a test — the tooltip fires at mount + 10 s and only sometimes intercepts, so no deterministic mutation proves it. It rests on #1220's measurement (a 282×32 px opaque element at z-index: 40, pointer-events: auto, over the canvas-controls bar) plus the fact that this spec is still clicking that bar past the 10 s mark.

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Updates the “auto-save off” regression coverage to use the shared openFlowById navigation helper and proactively seed the assistant onboarding flag to prevent UI interference/flakiness during repeated re-opens.

Changes:

• Replace direct page.goto('/flow/{id}') + selector wait with openFlowById, while preserving an explicit wait for GET /api/v1/flows/{id} to ensure the graph is applied before node-count assertions.
• Seed assistant onboarding state at the start of the test to avoid late-appearing overlays blocking adjustScreenView interactions.
• Refresh the accompanying doc to reflect the new helper-based re-open flow and the rationale for the added gating.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
tests/tests-automations/regression/flow-functionality/auto-save-off.spec.ts	Uses openFlowById for consistent re-open behavior and seeds onboarding early to reduce UI flake.
docs/flow-functionality/auto-save-off.md	Documents the helper-based re-open path and the onboarding seeding rationale.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.