dashboard

# `/dashboard` — GKE Capacity & Triage Dashboard Launch Antigravity counterpart of the Claude Code `/dashboard` command (`.claude/commands/dashboard.md`). Both configure the local dashboard daemon, seed monitored clusters in `_dashboard/clusters_config.json`, and generate deep-linked diagnostic URLs. ## Command Usage - `/dashboard [project_id]` (e.g. `/dashboard my-gcp-project`) - `/dashboard [cluster_name]` (e.g. `/dashboard my-tpu-cluster`) - `/dashboard [workload_id]` (e.g. `/dashboard my-tpu-workload`) - `/dashboard` (defaults to active gcloud project or ambient configuration) > [!NOTE] The local web UI daemon (`_dashboard/server.py`) is an optional > companion component. In external checkouts where `_dashboard/` is not bundled, > use `/troubleshoot` and `_wiki/tools/scripts/python_glue_runner.py` directly. ## Execution Flow & Actions 1. **Target Identification & Scoping**: - If a `project_id` is supplied, all active GKE clusters within the project are discovered and populated into `_dashboard/clusters_config.json`. - If a `cluster_name` is supplied, its location, generation, and reservation are resolved and added to `_dashboard/clusters_config.json`. - If a `workload_id` is supplied, the underlying cluster and region are resolved via `[[validate-workload]]` before configuring the dashboard. - If omitted, the active gcloud project or existing configured clusters are monitored. 2. **Server Health & Background Daemon**: - Verifies whether `http://localhost:8080/api/health` responds 200 OK. - If not running, launches `_dashboard/server.py --port 8080 --no-browser` as a detached background daemon and waits for health readiness. - If the port is held by a server that does not answer, reports `server_status: UNRESPONSIVE`; stop that process before retrying. - A newly started daemon runs AI triage on the launching harness (Antigravity here). The result's `agent_backend` names the backend the running server uses. If `agent_backend_warning` is set, a reused server uses the other harness: stop it and relaunch. 3. **Live Diagnostics Scan**: - Triggers `POST /api/scan` to immediately initiate a diagnostic scan across the monitored clusters. 4. **Deep-Linked URL Generation**: - Scopes the browser URL with query parameters (`cluster`, `workload`, `gen`, or `project`). - Deep-linking triggers automatic table filtering in `#workload-search` on the frontend. 5. **URL Verification (`[[verify-dashboard-url]]`)**: - Resolves the URL's hostname; an unresolvable FQDN (e.g. `*.local`) falls back to `localhost`. - Probes both the deep-linked page and `/api/health` on that host several times while the scan runs. A single health check is not enough: the server can answer once and then freeze. - Never report a URL whose `url_status` is not `VERIFIED` or `DEGRADED`; report `DEGRADED` with a latency caveat (isolated slow probes under host load). On `UNRESPONSIVE`, follow `url_hint`, restart the server, and re-verify with: ```bash python3 _wiki/tools/scripts/configure_and_launch_dashboard.py \ --verify-url '<url>' ``` ## Primary Workflow & Actions - `[[launch-workload-dashboard]]` - `[[configure-and-launch-dashboard]]` - `[[gcloud-login]]` - `[[resolve-cluster-metadata]]` - `[[validate-workload]]` - `[[verify-dashboard-url]]` ## Related Playbooks - `[[general-ml-debugging]]` - `[[workload-disruption-orchestration]]` ## Execution Command ```bash python3 _wiki/tools/scripts/configure_and_launch_dashboard.py "$@" ``` -------------------------------------------------------------------------------- ## Main Coordinator Delegation Invariant The main coordinator agent acts strictly as a high-level orchestrator and has **zero authorization to execute standalone shell commands, direct script invocations, log queries, file edits, or GCS uploads directly**. All execution MUST be delegated to the designated specialized wiki subagents via `invoke_subagent` / `send_message`: - **Workflow & Playbook Routing**: Delegate to `wiki_router`. - **Diagnostic & Action Execution**: Delegate to `wiki_practitioner` (for `/troubleshoot` milestones, Cloud Logging, GCS bucket scaffolding, empirical packaging, uploads, and background monitoring). - **Hypothesis & Gap Evaluation**: Delegate to `wiki_advisor`. - **Adversarial Verification & Audit**: Delegate to `wiki_reviewer` (for the `ai-verify-failure-attribution` prompt task before publication). - **Code, Script & Schema Edits**: Delegate to `wiki_maintainer`. The main agent manages user approvals, coordinates subagent dispatches, and presents final synthesized reports.