[GREEN] gflow nightly canary — 2026-11-14 02:00 UTC

This issue is a dashboard, not a bug report. Please don't close it.

A nightly canary runs the $0 live test tiers (-m e2e_auth) against real Google Flow
from the maintainer's machine, then rewrites this issue's title with the result.
The comments below are the run history.

Hosted CI can't do this: the live tiers need a real authenticated Chrome profile, and
Google bot-detection plus reCAPTCHA make hosted auth infeasible. So the run is local and
reports here instead. See #502 for the full rationale.

Reading the title

State	Meaning	Action needed
`[GREEN]`	all selected $0 tiers passed	none
`[RED]`	auth was healthy but a tier failed	real drift or regression — triage
`[AUTH-EXPIRED]`	the maintainer's Google session expired	re-login; not a code problem
`[DEFERRED]`	the profile was in use by another process	none — nothing ran

AUTH-EXPIRED and DEFERRED are deliberately kept out of RED, so that RED always
means "code or Flow drifted" — never "please re-login" or "you had a browser open".

What this does not do

Gates nothing. It blocks no PR and no release. A gate on a machine that might be
switched off would be self-inflicted downtime.
Never opens new issues. Every run edits this one. A new issue per night would be
365 a year, and nobody reads those.
Never spends credits. Only the zero-credit tiers run unattended. The credit tiers
(e2e_image / e2e_video / smoke) stay strictly manual.

Staleness is a signal

The title carries the last run's timestamp. If it's days old, the machine was off — the
canary is stale, not green. That ambiguity is why this reports on success as well as
failure, unlike a CI job whose run history speaks for itself.

Published content

Sanitized for a public repo: commit SHA, pass/fail/skip counts, duration, failure class,
and failing test names. Never raw logs, profile paths, prompts, or signed URLs.

Runner: scripts/canary/run_canary.py · docs: docs/E2E_TESTING.md § Nightly canary · tracked by #502

GREEN — 2026-08-20 21:03 UTC

All selected $0 tiers passed.

commit	`7abdb1b`
markers	`e2e_auth`
passed	0
failed	0
skipped	0
duration	0.0s

☝️ The entry above was a wiring test (--simulate GREEN), not a real run — note passed: 0, duration: 0.0s. It confirms the canary can edit this issue's title and post here. Title reset to [PENDING] until the first scheduled run.

RED — 2026-08-20 21:05 UTC

A $0 tier failed while auth was healthy — real drift or regression.

commit	`7abdb1b`
markers	`e2e_auth`
passed	17
failed	1
skipped	1
duration	153.2s

Failing:

tests.e2e.test_aisandbox_auth_live::test_rest_upload_image_authenticates_after_sapisidhash

Canary gates nothing. Triage at your convenience.

GREEN — 2026-08-21 06:15 UTC

All selected $0 tiers passed.

commit	`64d735c`
markers	`e2e_auth`
passed	18
failed	0
skipped	1
duration	150.7s

RED — 2026-08-23 06:14 UTC

A $0 tier failed while auth was healthy — real drift or regression.

commit	`230200b`
markers	`e2e_auth`
passed	17
failed	1
skipped	1
duration	135.8s

Failing:

tests.e2e.test_aisandbox_auth_live::test_rest_upload_image_authenticates_after_sapisidhash

Canary gates nothing. Triage at your convenience.

RED — 2026-08-25 02:00 UTC

A $0 tier failed while auth was healthy — real drift or regression.

commit	`230200b`
markers	`e2e_auth`
passed	17
failed	1
skipped	1
duration	171.6s

Failing:

tests.e2e.test_aisandbox_auth_live::test_rest_upload_image_authenticates_after_sapisidhash

Canary gates nothing. Triage at your convenience.

chore(canary): keep the captured logs on a RED, and correct the 401's…

RED — 2026-11-07 02:00 UTC

A $0 tier failed while auth was healthy — real drift or regression.

commit	`0cc97c20`
markers	`e2e_auth`
passed	65
failed	1
skipped	39
duration	2431.9s

Failing:

tests.e2e.test_missing_project_e2e::test_e2e_cli_t2i_into_a_missing_project_exits_11_fast

Unchanged for 7 consecutive runs. The same tests failing the same way is the shape of a standing condition — a capability this account lost, or a known-broken surface — not of fresh drift. If an issue already tracks it, link that issue here and stop re-triaging it nightly. If none does, it needs one: a red nobody owns is how a canary trains red-blindness.

RED — 2026-11-08 02:00 UTC

A $0 tier failed while auth was healthy — real drift or regression.

commit	`fbabf435`
markers	`e2e_auth`
passed	63
failed	3
skipped	63
duration	2583.7s

Failing:

tests.e2e.test_missing_project_e2e::test_e2e_cli_t2i_into_a_missing_project_exits_11_fast
tests.e2e.test_project_create_e2e::test_video_without_a_project_reaches_the_submit
tests.e2e.test_project_create_e2e::test_mcp_video_without_a_project_reaches_the_submit

The failing set changed since the last run. Whatever else is standing, something here is new — this is the arm worth looking at today.

RED — 2026-11-09 02:00 UTC

A $0 tier failed while auth was healthy — real drift or regression.

commit	`ca069fca`
markers	`e2e_auth`
passed	65
failed	1
skipped	63
duration	2382.2s

Failing:

tests.e2e.test_missing_project_e2e::test_e2e_cli_t2i_into_a_missing_project_exits_11_fast

The failing set changed since the last run. Whatever else is standing, something here is new — this is the arm worth looking at today.

RED — 2026-11-10 02:00 UTC

A $0 tier failed while auth was healthy — real drift or regression.

commit	`ca069fca`
markers	`e2e_auth`
passed	65
failed	1
skipped	63
duration	2491.0s

Failing:

tests.e2e.test_missing_project_e2e::test_e2e_cli_t2i_into_a_missing_project_exits_11_fast

Unchanged for 2 consecutive runs. The same tests failing the same way is the shape of a standing condition — a capability this account lost, or a known-broken surface — not of fresh drift. If an issue already tracks it, link that issue here and stop re-triaging it nightly. If none does, it needs one: a red nobody owns is how a canary trains red-blindness.

GREEN — 2026-11-11 02:00 UTC

All selected $0 tiers passed.

commit	`b7fc3b3f`
markers	`e2e_auth`
passed	75
failed	0
skipped	63
duration	1262.6s

RPC-DRIFT OK

Every catalogued RPC id is in the served JavaScript.

build	`boq_labs-ai-sandbox-frontend_20261110.00_p0`
catalogued ids found	28/28
descriptors	267

DEFERRED — 2026-11-12 02:00 UTC

Nothing conclusive ran. Neutral — this says nothing about Flow.

commit	`6a6ec430`
markers	`e2e_auth`
passed	0
failed	0
skipped	0
duration	0.0s

Reason: uv sync failed after pull; dependency tree may be stale

RPC-DRIFT OK

Every catalogued RPC id is in the served JavaScript.

build	`boq_labs-ai-sandbox-frontend_20261111.00_p0`
catalogued ids found	28/28
descriptors	268

New build since boq_labs-ai-sandbox-frontend_20261110.00_p0: 2 descriptor ids added, 1 removed.

added: PzXMtd, ll11Cf
removed: BN3L2e

GREEN — 2026-11-13 02:00 UTC

All selected $0 tiers passed.

commit	`96339be1`
markers	`e2e_auth`
passed	75
failed	0
skipped	63
duration	1189.9s

RPC-DRIFT OK

Every catalogued RPC id is in the served JavaScript.

build	`boq_labs-ai-sandbox-frontend_20261111.08_p0`
catalogued ids found	28/28
descriptors	267

New build since boq_labs-ai-sandbox-frontend_20261111.00_p0: 0 descriptor ids added, 1 removed.

removed: CsdYNb

GREEN — 2026-11-14 02:00 UTC

All selected $0 tiers passed.

commit	`96339be1`
markers	`e2e_auth`
passed	75
failed	0
skipped	63
duration	988.4s

RPC-DRIFT OK

Every catalogued RPC id is in the served JavaScript.

build	`boq_labs-ai-sandbox-frontend_20261111.08_p0`
catalogued ids found	28/28
descriptors	267

Navigation Menu

Repository navigation

[GREEN] gflow nightly canary — 2026-11-14 02:00 UTC #559

Reading the title

What this does not do

Staleness is a signal

Published content

GREEN — 2026-08-20 21:03 UTC

RED — 2026-08-20 21:05 UTC

GREEN — 2026-08-21 06:15 UTC

RED — 2026-08-23 06:14 UTC

RED — 2026-08-25 02:00 UTC

159 remaining items

RED — 2026-11-07 02:00 UTC

RED — 2026-11-08 02:00 UTC

RED — 2026-11-09 02:00 UTC

RED — 2026-11-10 02:00 UTC

GREEN — 2026-11-11 02:00 UTC

RPC-DRIFT OK

DEFERRED — 2026-11-12 02:00 UTC

RPC-DRIFT OK

GREEN — 2026-11-13 02:00 UTC

RPC-DRIFT OK

GREEN — 2026-11-14 02:00 UTC

RPC-DRIFT OK

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

[GREEN] gflow nightly canary — 2026-11-14 02:00 UTC #559

Description

Reading the title

What this does not do

Staleness is a signal

Published content

Activity

ffroliva commented on Aug 20, 2026

GREEN — 2026-08-20 21:03 UTC

ffroliva commented on Aug 20, 2026

ffroliva commented on Aug 20, 2026

RED — 2026-08-20 21:05 UTC

ffroliva commented on Aug 21, 2026

GREEN — 2026-08-21 06:15 UTC

ffroliva commented on Aug 23, 2026

RED — 2026-08-23 06:14 UTC

ffroliva commented on Aug 25, 2026

RED — 2026-08-25 02:00 UTC

159 remaining items

ffroliva commented on Nov 7, 2026

RED — 2026-11-07 02:00 UTC

ffroliva commented on Nov 8, 2026

RED — 2026-11-08 02:00 UTC

ffroliva commented on Nov 9, 2026

RED — 2026-11-09 02:00 UTC

ffroliva commented on Nov 10, 2026

RED — 2026-11-10 02:00 UTC

ffroliva commented on Nov 11, 2026

GREEN — 2026-11-11 02:00 UTC

RPC-DRIFT OK

ffroliva commented on Nov 12, 2026

DEFERRED — 2026-11-12 02:00 UTC

RPC-DRIFT OK

ffroliva commented on Nov 13, 2026

GREEN — 2026-11-13 02:00 UTC

RPC-DRIFT OK

ffroliva commented on Nov 14, 2026

GREEN — 2026-11-14 02:00 UTC

RPC-DRIFT OK

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions