# Runs workspace

Purpose
- Monitor Scheduler runs, progress, deltas, and evidence bundles in real time.

Access and dependencies
- Route: /console/runs with /console/runs/:runId detail drawer.
- Scopes: runs.read; runs.manage for cancel or retry; policy:runs; downloads.read.
- Depends on Scheduler WebService, Policy Engine run summaries, Scanner evidence endpoints.
- Feature flags: runs.dashboard.enabled, runs.sse.enabled, runs.retry.enabled, runs.evidenceBundles.

Layout
- Header with tenant badge, schedule selector, backlog metrics.
- Cards for active runs, queue depth, new findings, KEV deltas.
- Tabs: active, completed, scheduled, failures.
- Runs table with detail drawer for summary, segments, deltas, evidence, logs.

Runs table
- Run ID: deterministic run:<tenant>:<timestamp>:<nonce>.
- Trigger: cron, manual, concelier, excititor, policy, content-refresh.
- State: planning, queued, running, completed, cancelled, error with ERR_RUN codes.
- Progress: processed vs total, updated via SSE.
- Duration: elapsed or total duration.
- Deltas: severity deltas and KEV changes.
- Filters: trigger, state, schedule, severity impact, policy revision, timeframe, shard, error code.

Detail drawer
- Summary: tenant, trigger, schedule, shard count, timestamps, correlation ID.
- Progress: segmented bar (planner, queue, execution, post-processing).
- Segments: retry failed segments with runs.manage.
- Deltas: links back to Findings filtered by run ID.
- Evidence: policy run summary, findings delta CSV, scanner bundle, DSSE links.
- Logs: latest structured logs with correlation IDs and log pivot copy.

Queue and schedule management
- Side panel lists upcoming schedules with cron, timezone, enable toggles.
- Preview impact estimates candidate counts before launch.
- Manual run form supports analysis-only and content-refresh modes.
- Pause and resume schedules with confirmation.

Live updates
- SSE endpoint /console/runs/{id}/stream with stateChanged, segmentProgress, deltaSummary, log.
- UI reconnects with exponential backoff and heartbeat.
- Offline mode disables SSE and uses polling.

Retry and remediation
- Segment-level retry with cooldown timers.
- Full retry creates a new run ID with retryOf reference.
- Escalation template includes run context and correlation IDs.
- CLI parity: stella runs retry --run <id>, stella runs cancel --run <id>.

Evidence downloads
- Evidence tab aggregates run evidence bundles and manifest hashes.
- Bundle for offline packages all evidence into a single tarball.
- Completed bundles appear in the Downloads workspace.

Observability
- Metrics: scheduler_queue_depth, scheduler_runs_active, scheduler_runs_error_total, scheduler_runs_duration_seconds.
- Trend charts for queue depth, runs per trigger, determinism score.
- Alerts for planner lag, queue saturation, repeated error codes.

Offline behavior
- Snapshot banner shows staleness and disables SSE.
- Manual run form produces CLI scripts for offline execution.
- Evidence downloads default to local paths for transfer.
- Tenants missing in the snapshot are hidden.