Add reliable execution reporting and coordinated dashboard views

scripts/report_relay.py

@@ -0,0 +1,32 @@
+#!/usr/bin/env python3
+"""Retry only persisted idempotent status reports, never repeat model work."""
+import argparse
+import json
+from pathlib import Path
+import time
+from report import flush
+from worker import atomic_json,session_lock
+
+
+def main():
+    p=argparse.ArgumentParser(description=__doc__)
+    p.add_argument('--config',required=True)
+    p.add_argument('--once',action='store_true')
+    a=p.parse_args()
+    bridge=json.loads(Path(a.config).read_text(encoding='utf-8-sig'))
+    cfg=json.loads(Path(bridge['task_report_config']).read_text(encoding='utf-8-sig'))
+    health=Path(bridge['state_file']).parent/'report_relay-health.json'
+    with session_lock('task-report-relay:'+json.dumps(cfg,sort_keys=True)):
+        while True:
+            try:
+                counts=flush(cfg)
+                atomic_json(health,{'status':'ok' if not counts['pending'] and not counts['blocked'] else 'failed',
+                                   'updated':time.time(),'reports':counts})
+            except Exception:
+                atomic_json(health,{'status':'failed','updated':time.time(),'reason':'Inspect private report outbox'})
+                if a.once:raise
+            if a.once:return
+            time.sleep(5)
+
+
+if __name__=='__main__':main()