CI failure: Test Code Samples

[github-actions]

The scheduled job Test Code Samples failed.

• Workflow: Test Code Samples
• Failing run: https://github.com/viamrobotics/docs/actions/runs/28525092525

Please investigate: open a fix PR if the cause is clear, otherwise
comment on this issue explaining what is needed. This issue is
created automatically on failure and reused (commented on) for
repeat failures.

[btshrewsbury-viam]

Triage (from the branch dispatch in #5120, which isolates the language steps so all failures surface at once):

1. Python fleet-api/fleet-management-api-orgs.py — AssertionError: Expected at least one organization member without existing authorizations. Needs a roleless member in the docs-scheduled-tests test org (data dependency, not a code bug).
2. Go fleet-api/fleet-management-api-orgs-metadata.go:154 — OrganizationSetSupportEmail(ctx, ORG_ID, "none") is now rejected: InvalidArgument: invalid email address format. Sample bug.
3. TypeScript capture-annotate-images/capture-images.ts:71 and capture-annotate-dataset.ts:91 — TypeError: Cannot read properties of undefined (reading 'getTime') in binaryDataCaptureUpload (@viamrobotics/sdk ^0.55.0). Sample bug (signature drift).

#5120 fixes the notification + step-isolation infrastructure; these four sample fixes are separate.

[github-actions]

Job Test Code Samples failed again. Run: https://github.com/viamrobotics/docs/actions/runs/28527066008

[github-actions]

Job Test Code Samples failed again. Run: https://github.com/viamrobotics/docs/actions/runs/28529469071

[btshrewsbury-viam]

Update on orgs.py (Python): the migration to a reachable org (docs-scheduled-tests) got it past the member/invite/role selection and a location-count assertion, but surfaced a design problem, not a config gap:

• Viam has no such thing as a role-less org member. Removing a member's last authorization removes them from the org. (Confirmed empirically: stripping the test member's only role dropped the member count from 2 to 1.)
• fleet-management-api-orgs.py both requires a member that starts with zero authorizations and ends by remove_role, which returns that member to zero authorizations — i.e. it consumes its own test member every run. So even a hand-seeded roleless member can't survive one successful run.

This is very likely the root cause of the job failing since 2025-11-10; #5106's "select a roleless member" change can't be satisfied in steady state.

Fixing this requires redesigning the role-lifecycle section so it doesn't require/consume a zero-auth member (e.g. operate on a member that keeps a baseline operator role, demonstrating add/change/remove of an additional role), or excluding this sample from the live suite. Tracking here for a decision.

Everything else in the suite now passes (Go and TS code bugs fixed; the occasional TS "Failed to connect to robot" is transient machine flakiness).

[github-actions]

Job Test Code Samples failed again. Run: https://github.com/viamrobotics/docs/actions/runs/28531956095

[btshrewsbury-viam]

orgs.py is fixed (#5120): redesigned to operate on a non-owner (operator) member and add/move/remove an additional role, so the member is never consumed. Verified passing in CI and repeatably against the dedicated test org.

Remaining test-code-snippets failures are infrastructure flakiness, not code:

• Machine-dependent TS samples intermittently fail with "Failed to connect to robot" (the shared cloud test machine is frequently unreachable — offline in 2 of the last 3 runs).
• query/data-query.py intermittently fails with DEADLINE_EXCEEDED (backend SQL query timeout).

All actual sample code bugs are resolved. Stabilizing the rest requires a reliable test machine (the current shared one is not administrable by us) and/or tolerance for transient backend timeouts.

[github-actions]

Job Test Code Samples failed again. Run: https://github.com/viamrobotics/docs/actions/runs/28538719942

[github-actions]

Job Test Code Samples failed again. Run: https://github.com/viamrobotics/docs/actions/runs/28538924113

[btshrewsbury-viam]

Checked the 3 runs since the last update (28531956095, 28538719942, 28538924113) directly against their job logs. All code-bug fixes from #5120 continue to hold (orgs.py, orgs-metadata.go, and the TS capture samples all pass consistently). Each of these 3 runs failed on a different sample with a transient, non-reproducible error:

• 28538719942: Go run-inference/run-inference.go — resource rdk:component:camera/camera-1 not found (shared test machine's resource config state)
• 28538719942: Go data-pipelines/pipeline-disable.go — rpc error: code = Internal desc = ...an internal error occurred (backend transient)
• 28538924113: Python capture-annotate-images/label-images.py — ConnectionError: Unable to establish a connection to the machine (shared test machine unreachable)

No consistent single file is failing, and none of these are code bugs — all point to the same shared-test-machine/backend flakiness already flagged. No fix possible via a docs edit; leaving open as infrastructure tracking per the prior triage.

---

Generated by Claude Code