Search Google Cloud (GKE) server logs of the DEV environment (dev.voxt.ai). Use when troubleshooting anything that needs the server side of the story — stream stalls, RPC peer changes / reconnects, watchdog closures, mesh/node churn, pod restarts — or when the user says "check GCP logs", "check server logs on dev", or gives a dev.voxt.ai timeline from a client console log.
The English summary is being prepared. Showing the original description.
Review the instructions the agent will receive before you install.
actual-chat-app-dev (usually already the active gcloud config project).resource.labels.container_name="actual-chat-app".gcloud auth list must show an active account. If not, ask the user to run
! gcloud auth login — never attempt it yourself.gcloud logging read \
'resource.labels.container_name="actual-chat-app"
AND timestamp>="2026-07-22T05:12:00Z" AND timestamp<="2026-07-22T05:17:00Z"
AND textPayload:"<term>"' \
--project=actual-chat-app-dev \
--format='value(timestamp, textPayload)' \
--order=asc | head -80
timestamp on both ends; unbounded queries are slow and get truncated.
Start with a ±3-5 min window around the incident and widen only if needed.textPayload:"x" is a substring match; combine with AND/OR in parentheses.--order=asc + head gives a readable timeline; default order is desc.resource.labels.pod_name to --format='value(...)' when multiple pods matter.| Looking for | Term |
|---|---|
| A specific media stream | short stream-id fragment, e.g. "XDWKB" (ids look like <nodeId>-01KY43XDWKB0…; the prefix before -01… is the serving NODE id, not the author) |
| Publish start | "PushStream: VideoRecord" / "PushStream: AudioRecord" |
| Server killed a silent stream | "Stale stream watchdog" |
| Stream registration / viewers | "RegisterActiveStream", "GetStream: first frame", "GetVideoRaw" |
| Demand fan-out | "SubscribeToDemand", "DemandInfo" |
| Mesh / pod churn (client peer-change trigger) | "UpdateOnlineNodes", "ComputeState", "Rerouting", "Shards @" |
| Keyframe requests | "RequestKeyFrame" |
PushStream, watchdog, UpdateOnlineNodes) in the same window.UpdateOnlineNodes/Rerouting = node churn — expect
client-side peerChanged=true reconnects at the same moment.No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Delete the `-bak` backup branches that `/prepare-merge` leaves behind, local and on origin, once the work they back up has merged. Use when the user says "remove my bak branches", "clean up bak branches", "delete the backups", or "/clean-bak-branches".
The English summary is being prepared. Showing the original description.
Use when work on a branch is finished and the next step is a pull request — "create a PR", "open a PR", "push and PR this", "/create-pr" — when a draft PR should be marked ready for review — "mark the PR ready", "undraft it", "it's ready for review" — or right after either has happened and the team has not been told about it yet.
The English summary is being prepared. Showing the original description.
Search Google Cloud (GKE) server logs of the PROD environment. Use when troubleshooting a production incident that needs the server side of the story — stream stalls, RPC peer changes / reconnects, watchdog closures, mesh/node churn, pod restarts — or when the user says "check PROD logs", "check production server logs", or gives a production timeline from a client console log.
The English summary is being prepared. Showing the original description.
What this project has already learned the hard way, and where any new such fact belongs. Two jobs. (1) Routing: this project keeps NO per-project Claude memories — read this before writing anything down, and pick a durable home from the list here. (2) An index, by area, of non-obvious facts worth knowing before you guess: debugging rigs and measurements that lie, iOS/WebKit and the tethered iPhone, the Windows app, production and deployment, and general codebase traps. Use before setting up a browser- or server-driven debugging session, when a bug reproduces only on iPhone, in Safari or in the Windows app, before changing deployment configuration or reading store/crash data, when something behaves unexpectedly and you are about to guess why, and whenever you are tempted to record a fact for later.
The English summary is being prepared. Showing the original description.
Review the EF Core migrations a release (or a PR) carries against how Voxt is deployed: a rolling update whose new pods wait for the migration Job, while old pods keep serving. Use before cutting a release branch, before merging a PR that adds a migration, or when the user says "check the migrations", "are the migrations safe to ship", "expand/contract check", "/migration-release-check". Reports risks; never blocks, edits or runs anything on prod.
The English summary is being prepared. Showing the original description.
CEO/founder-mode plan review. Rethink the problem, find the 10-star product, challenge premises, expand scope when it creates a better product. Three modes: SCOPE EXPANSION (dream big), HOLD SCOPE (maximum rigor), SCOPE REDUCTION (strip to essentials).
The English summary is being prepared. Showing the original description.