# Automation Page (AI Automation Pipeline Dashboard)
 
 ## Purpose
 Provide a site-scoped dashboard to configure, run, pause, resume, and monitor the 7-stage automation pipeline. Surfaces pipeline overview, current run status, metrics, history, configuration modal, and credit sufficiency checks.
 
 ## Code Locations (exact paths)
 - Primary page: `frontend/src/pages/Automation/AutomationPage.tsx`
 - Service: `frontend/src/services/automationService` (config, current run, estimate, pipeline overview, runNow, pause, resume, publishWithoutReview)
 - Metrics sources: `frontend/src/services/api` (`fetchKeywords`, `fetchClusters`, `fetchContentIdeas`, `fetchTasks`, `fetchContent`, `fetchContentImages`)
 - UI components: `frontend/src/components/Automation/{ActivityLog,ConfigModal,RunHistory,CurrentProcessingCard}`, `frontend/src/components/dashboard/EnhancedMetricCard`, `frontend/src/components/common/ComponentCard`, `frontend/src/components/common/PageMeta`, `frontend/src/components/common/DebugSiteSelector`
 
 ## High-Level Responsibilities
 - Load automation config, current run, credit estimate, and pipeline overview for the active site.
 - Poll current run and pipeline status while running/paused; refresh metrics regularly during runs.
 - Provide controls: Run Now, Pause, Resume, Save Config, Publish Without Review.
 - Show per-stage cards, current processing card, run history, and activity log.
 
 ## Detailed Behavior
 - Site binding: requires `useSiteStore.activeSite`; without it, page shows a “select a site” message.
 - Initial load (`loadData`): parallel calls to `getConfig`, `getCurrentRun`, `estimate`, `getPipelineOverview` plus low-level metrics (keywords/clusters/ideas/tasks/content/images counts) per site_id.
 - Polling: 5s interval; if run status is `running`/`paused`, refresh run, pipeline, and metrics; otherwise refresh pipeline only.
 - Metrics: counts for keywords (total/new/mapped), clusters (total/new/mapped), ideas (total/new/queued/completed), tasks (total), content (total/draft/review/published), images (total/pending).
 - Stage cards: derived from `STAGE_CONFIG` array representing 7 pipeline stages including manual review gate.
 - Actions:
   - Run Now: checks credit sufficiency from `estimate`; blocks if insufficient.
   - Pause/Resume: call automationService with site + run_id, then refresh run/pipeline/metrics.
   - Save Config: persists partial config, updates local state, reloads pipeline/metrics/run.
   - Publish Without Review: calls `publishWithoutReview` with confirmation prompt, then reloads.
 - UI states: `loading` spinner until initial fetch; `showProcessingCard` toggled on when a run exists.
 
 ## Data Structures / Models Involved (no code)
 - `AutomationConfig`: site-level automation settings (intervals, gates, etc.).
 - `AutomationRun`: run_id, status (`running`, `paused`, etc.), stage info.
 - `PipelineStage`: stage list with status/progress.
 - Metrics DTOs from planner/writer/image endpoints (`count` fields only).
 
 ## Execution Flow
 - `useEffect` with site dependency → `loadData`.
 - Interval polling while active runs → `loadCurrentRun`, `loadPipelineOverview`, `loadMetrics`.
 - User actions dispatch to automationService; toasts for success/error; follow-up refreshes.
 
 ## Cross-Module Interactions
 - Pulls planner/writer/image counts to present authoritative pipeline context.
 - Uses `useSiteStore` for tenant/site scoping; credit checks rely on billing estimate API.
 
 ## State Transitions
 - `currentRun.status` drives polling and control availability.
 - `showProcessingCard` turns on when a run exists.
 - `estimate.sufficient` gates Run Now action.
 
 ## Error Handling
 - Toast-based error reporting; fetch failures log to console but keep page usable.
 - Metrics fetch wrapped in try/catch; failure degrades to missing metrics without blocking the rest.
 
 ## Tenancy Rules
 - All automation service calls require `activeSite.id`; backend enforces account/site scoping. No client override for other tenants.
 
 ## Billing Rules (if applicable)
 - Run Now is blocked when `estimate.sufficient` is false; shows required vs current credits.
 
 ## Background Tasks / Schedulers (if applicable)
 - Client-side polling (5s) during active runs; no background scheduling beyond that.
 
 ## Key Design Considerations
 - Polling keeps UI aligned with long-running Celery pipeline states.
 - Credit gate prevents user-initiated runs that would immediately fail server-side.
 - Metrics are read-only mirrors of backend aggregates to align UI with planner/writer state.
 
 ## How Developers Should Work With This Module
 - When adding new stages, extend `STAGE_CONFIG` and ensure backend pipeline overview includes them.
 - Keep polling interval modest; if adding heavier metrics, consider staggering fetches to avoid rate limits.
 - Wire new config fields through `AutomationConfig` type, `ConfigModal`, and `updateConfig` payload.