Deploy GitHub main a3d79b5cbaf5

cloud/web.py

@@ -20,6 +20,9 @@ def health_report(root: Path = ROOT):
     supervisor = read(".cloud-state/supervisor.json")
     checkpoint = read(".cloud-state/checkpoint.json")
     status = read("space/results/online_status.json")
+    publication = read("space/results/publication_status.json") or status
+    if status.get("push_status") == "failed":
+        publication = status
     stamp = checkpoint.get("checkpointed_at")
     def age_seconds(stamp):
         try:
@@ -27,13 +30,13 @@ def health_report(root: Path = ROOT):
         except (ValueError, TypeError):
             return None
     age = age_seconds(stamp)
-    push_age = age_seconds(status.get("pushed_at"))
+    push_age = age_seconds(publication.get("pushed_at"))
     failures = status.get("failed_forecasts", [])
     model_failures = sum(item.get("model") != "*" for item in failures)
     healthy = bool(supervisor.get("worker_running") and supervisor.get("evaluator_enabled")
                    and 0 <= time.time() - supervisor.get("updated_at", 0) < 90
                    and age is not None and 0 <= age < 3600
-                   and status.get("aggregate_status") == "ok" and status.get("push_status") == "ok"
+                   and status.get("aggregate_status") == "ok" and publication.get("push_status") == "ok"
                    and push_age is not None and 0 <= push_age < 3600 and model_failures == 0)
     return {"ok": healthy, "source_revision": supervisor.get("source_revision"),
             "evaluator_enabled": supervisor.get("evaluator_enabled", False),
@@ -44,7 +47,7 @@ def health_report(root: Path = ROOT):
             "model_failure_count": model_failures,
             "unavailable_source_count": len(failures) - model_failures,
             "pending_tasks": status.get("pending_tasks"),
-            "push_status": status.get("push_status"), "pushed_at": status.get("pushed_at")}
+            "push_status": publication.get("push_status"), "pushed_at": publication.get("pushed_at")}
 
 
 def main():

deployment.json

@@ -1,5 +1,5 @@
 {
   "source_repository": "https://github.com/Thinkcat-Lab/LiveHouse-TS",
   "source_branch": "main",
-  "source_revision": "8c9fc27f33a5bfd1b3e4c4d1ff5e7d8c2aa05361"
+  "source_revision": "a3d79b5cbaf5ece22b8ad6dc4338b2a2a0515f81"
 }

scripts/push_results_to_hf.py

@@ -148,6 +148,12 @@ def update_push_status(results: Path, *, ok: bool, detail: str = "") -> None:
     if detail:
         payload["push_detail"] = detail
     status_path.write_text(json.dumps(payload, indent=4) + "\n")
+    # Evaluation rewrites online_status while rapid cycles wait for publication.
+    # Keep the last actual publish attempt independently for the watchdog.
+    receipt = results / "publication_status.json"
+    temporary = receipt.with_suffix(".tmp")
+    temporary.write_text(json.dumps({key: payload[key] for key in ("push_status", "pushed_at")}))
+    temporary.replace(receipt)
 
 
 def restart_space(token: str, repo_id: str) -> None: