Repository navigation
Deploy staging on every push to main. - #318
Conversation
Bugbot couldn't run - usage limit reachedBugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit. A user or team admin can review and increase usage limits in the Cursor dashboard. (requestId: serverGenReqId_c5e6a781-eb25-45b1-a465-a88d8bc26e6f) |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: bf049ea9f9
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Bugbot couldn't run - usage limit reachedBugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit. A user or team admin can review and increase usage limits in the Cursor dashboard. (requestId: serverGenReqId_10a78d8f-35ee-4cf2-b247-7e6ddec29bb1) |
Staging exists so a production deploy runs code that staging already ran. While staging deploys were manual only, staging went stale and the manual production deploy became the first run of that code. The push filter is the one platform-ci.yml uses. A `platform/**`-only filter would be too narrow: the workers bundle the SDK, so a change under `packages/**` must redeploy staging too. A `concurrency` group with `cancel-in-progress: false` keeps one deploy running and at most one pending, so a burst of pushes never interrupts a deploy between the api migration and the workers that depend on it. Also restore the smoke test's daily cron (`0 0 * * *`), which the old repo ran. Production stays on manual dispatch.
`platform/AGENTS.md` symlinks to `README.md`, which still described every platform deploy as manual, so the policy change in this PR left agents and operators with contradicting instructions. State which environment is manual and point at the workflow triggers, keeping the guardrail against deploying unprompted.
5f2888c to
ee924d5
Compare
Bugbot couldn't run - usage limit reachedBugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit. A user or team admin can review and increase usage limits in the Cursor dashboard. (requestId: serverGenReqId_600da5df-9521-44f9-b27e-cfd4ca2b7a83) |
The platform was merged into this repo under
platform/, and the first staging deploy frommainhas now run green (run 34808786846, dispatched by hand). This PR makes staging deploy on its own from then on.Why
Staging exists so a production deploy runs code that staging already ran. With manual-only staging deploys, staging goes stale and the manual production deploy becomes the first run of that code.
What changed
platform-deploy-staging.yml— deploy on push tomainThe path filter is the one
platform-ci.ymluses, deliberately: aplatform/**-only filter would be too narrow, because the workers bundle the SDK, so a change underpackages/**must redeploy staging too.mainhere is much busier than the old repo'smain(1–3 pushes/day there; here 1–11 normally and 70–90 on burst days, of which 174 of the last 179 pushes match this filter).cancel-in-progress: falsekeeps one deploy running and at most one pending, so a burst turns into a queue instead of a fleet of concurrent deploys, and a deploy is never interrupted between theapimigration and the workers that depend on it.platform-smoke-test.yml— restore the daily cron0 0 * * *(8am SGT), matching what the old repo ran.platform/meta/operations.md— describe the new trigger modelThe environments table, the secrets note, and the workflow list all said both deploy workflows run on manual dispatch only.
Production is untouched:
platform-deploy.ymlstays manual dispatch.Follow-ups once this merges
Smoke Testworkflow (void-platform), which still deploys old-repo kitchen-sink code to staging every night. Its staging deploy was already disabled.void-platform'sDeploy Productionstays active until production is cut over.Verified
actionlinton all platform workflows: cleanvp fmt --check: cleanNote
Medium Risk
Every matching
mainpush runs a full staging deploy (API migrations plus workers), so workflow or path-filter mistakes could redeploy staging frequently or at the wrong time; concurrency limits blast radius but not deploy frequency.Overview
Staging deploys automatically on
mainpushes that touch platform/SDK paths (same filter asplatform-ci.yml, includingpackages/**because workers bundle the SDK). Production stays manual-only viaplatform-deploy.yml.platform-deploy-staging.ymladds apushtrigger onmainwith that path list, plus aconcurrencygroup (platform-deploy-staging,cancel-in-progress: false) so busymainqueues deploys instead of cancelling mid-migration.platform-smoke-test.ymlregains a daily cron (0 0 * * *, 8am SGT) in addition to manual dispatch.Docs in
platform/README.mdandplatform/meta/operations.mdnow describe staging as auto-on-push and document the concurrency behavior; production trigger wording is unchanged.Reviewed by Cursor Bugbot for commit ee924d5. Bugbot is set up for automated code reviews on this repo. Configure here.