feat: Auto-retry failed proxies after cooldown period

- Add last_failed_at column to track failure time
- Failed proxies auto-retry after 4 hours (configurable)
- Proxies permanently failed after 10 failures
- Add /retry-stats and /reenable-failed API endpoints
- markProxySuccess() re-enables recovered proxies

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

backend/migrations/095_proxy_auto_retry.sql

@@ -0,0 +1,81 @@
+-- Migration: Auto-retry failed proxies after cooldown period
+-- Proxies that fail will be retried after a configurable interval
+
+-- Add last_failed_at column to track when proxy last failed
+ALTER TABLE proxies ADD COLUMN IF NOT EXISTS last_failed_at TIMESTAMP;
+
+-- Add retry settings
+INSERT INTO settings (key, value, description)
+VALUES
+  ('proxy_retry_interval_hours', '4', 'Hours to wait before retrying a failed proxy'),
+  ('proxy_max_failures_before_permanent', '10', 'Max failures before proxy is permanently disabled')
+ON CONFLICT (key) DO NOTHING;
+
+-- Create function to get eligible proxies (active OR failed but past retry interval)
+CREATE OR REPLACE FUNCTION get_eligible_proxy_ids()
+RETURNS TABLE(proxy_id INT) AS $$
+DECLARE
+  retry_hours INT;
+BEGIN
+  -- Get retry interval from settings (default 4 hours)
+  SELECT COALESCE(value::int, 4) INTO retry_hours
+  FROM settings WHERE key = 'proxy_retry_interval_hours';
+
+  RETURN QUERY
+  SELECT p.id
+  FROM proxies p
+  WHERE p.active = true
+     OR (
+       p.active = false
+       AND p.last_failed_at IS NOT NULL
+       AND p.last_failed_at < NOW() - (retry_hours || ' hours')::interval
+       AND p.failure_count < 10  -- Don't retry if too many failures
+     )
+  ORDER BY
+    p.active DESC,  -- Prefer active proxies
+    p.failure_count ASC,  -- Then prefer proxies with fewer failures
+    RANDOM();
+END;
+$$ LANGUAGE plpgsql;
+
+-- Create scheduled job to periodically re-enable proxies past their retry window
+-- This runs every hour and marks proxies as active if they're past retry interval
+CREATE OR REPLACE FUNCTION auto_reenable_proxies()
+RETURNS INT AS $$
+DECLARE
+  retry_hours INT;
+  max_failures INT;
+  reenabled_count INT;
+BEGIN
+  -- Get settings
+  SELECT COALESCE(value::int, 4) INTO retry_hours
+  FROM settings WHERE key = 'proxy_retry_interval_hours';
+
+  SELECT COALESCE(value::int, 10) INTO max_failures
+  FROM settings WHERE key = 'proxy_max_failures_before_permanent';
+
+  -- Re-enable proxies that have cooled down
+  UPDATE proxies
+  SET active = true,
+      updated_at = NOW()
+  WHERE active = false
+    AND last_failed_at IS NOT NULL
+    AND last_failed_at < NOW() - (retry_hours || ' hours')::interval
+    AND failure_count < max_failures;
+
+  GET DIAGNOSTICS reenabled_count = ROW_COUNT;
+
+  IF reenabled_count > 0 THEN
+    RAISE NOTICE 'Auto-reenabled % proxies after % hour cooldown', reenabled_count, retry_hours;
+  END IF;
+
+  RETURN reenabled_count;
+END;
+$$ LANGUAGE plpgsql;
+
+-- Add index for efficient querying
+CREATE INDEX IF NOT EXISTS idx_proxies_retry
+ON proxies(active, last_failed_at, failure_count);
+
+COMMENT ON COLUMN proxies.last_failed_at IS 'Timestamp of last failure - used for auto-retry logic';
+COMMENT ON FUNCTION auto_reenable_proxies() IS 'Call periodically to re-enable failed proxies that have cooled down';