Skip to content

Prompt enum dropdowns: read from codegen instead of runtime introspection - #466

Merged
saqibmanan merged 12 commits into
mainfrom
dev
Sep 16, 2026
Merged

saqibmanan merged 12 commits into
mainfrom
dev

Conversation

@saqibmanan

Copy link
Copy Markdown
Contributor

Promotes the enum fix from dev to prod. Closes the last known provider test failure.

What ships: #465 only — the five prompt/AI-model dropdowns now read the generated enums, so there is no __type introspection request to abort and no cached failure. Workflow files are byte-identical to what is already on main.

Evidence: frontend dev deploy 35057507316 — build, smoke 3/3, functional 2/2, report-only provider 11 passed 0 failed (test_prv_006b green for the first time), promote success.

Note: the Full Suite (dev) api-smoke check is expected to fail here — it verifies the dev site is running the PR head SHA, which a dev→main PR ref never matches.

Prod had no tests at all: a release was marked last-known-good straight
after deploy. Now main runs only the readonly allowlist against prod and
promotes on pass or rolls back on failure; dev runs smoke, then
functional only if smoke passed. api_base_url is explicit per branch --
run-smoke's api job otherwise falls back to the dev API.
Provider functional tests fail at shifting points against dev (autosave
races, slow create flows), so gating on them would roll back dev deploys
spuriously. The gating functional stage and the PR gate now skip
provider; separate provider jobs run with only_provider and report,
without being a dependency of rollback/promote or a required check.
Finalize and rollback did not list the report-only provider job in needs,
so a deploy asked for finalize approval while provider tests were still
running. Both now wait for it, and condition on the gating jobs' results
explicitly instead of failure()/!failure(), which would count the
report-only job's result.
…pproval

production requires a reviewer, and every job referencing it asked again:
approve the deploy, then approve finalize or rollback -- so a failing prod
deploy stayed live until someone clicked. production-ops holds the same host
and SSH secrets with no reviewer and a main-only branch policy; the human
approval stays on the deploy job.
feat: deploy gating model — readonly prod + rollback, two-stage dev, PR gate
…) [skip ci]

revert: keep prod finalize/rollback in the production environment
* fix: pass KEYCLOAK_CLIENT_SECRET to the report-only provider job

org_add_permission needs a Keycloak token to check canAdd. A called
workflow only sees secrets its caller passes, and the report-only job
passed none, so every org-create test skipped on both workers -- the
provider report looked green while never exercising org flows.

* fix: pass KEYCLOAK_CLIENT_SECRET to the report-only provider-report job

Same gap as the deploy workflow's report job: without it every
org-create test skips on both workers.
The four __type introspection queries fired together on form load; when the
backend was slow the browser aborted them and staleTime: Infinity cached the
failure, leaving Task Type / Domain / Target Languages / Target Model Types
empty until a reload. Enum values are fixed at build time, so read them from
the generated enums instead.
Same abort-and-cache failure as the metadata form; also drops a debug effect
that logged the whole enum payload to the console.
@saqibmanan saqibmanan self-assigned this Sep 16, 2026
@saqibmanan
saqibmanan merged commit 598c763 into main Sep 16, 2026
24 of 25 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant