From bbf83f02b2d10f67c994836acc57614bff6be43d Mon Sep 17 00:00:00 2001
From: arc53-machine <232052973+arc53-machine@users.noreply.github.com>
Date: Tue, 29 Sep 2026 10:40:23 +0100
Subject: [PATCH] Skip scheduled sync of a remote source paused for reconnect

S3 and GitHub sources synced from a connection kept being retried on
every schedule after their connection needed reconnecting, failing and
notifying the owner each time. They now wait, like Drive and Confluence
sources, and resume when the connection is reconnected.
---
 docsgpt/worker.py                | 19 ++++++++++++++++++-
 tests/connectors/test_runtime.py | 21 +++++++++++++++++++++
 2 files changed, 39 insertions(+), 1 deletion(-)

diff --git a/docsgpt/worker.py b/docsgpt/worker.py
index 5e70393c..0e5242bb 100755
--- a/docsgpt/worker.py
+++ b/docsgpt/worker.py
@@ -1519,7 +1519,7 @@ def sync_worker(self, frequency):
     with db_readonly() as conn:
         result = conn.execute(
             sql_text(
-                "SELECT id, name, user_id, type, remote_data, retriever, connection_id "
+                "SELECT id, name, user_id, type, remote_data, retriever, connection_id, metadata "
                 "FROM sources WHERE sync_frequency = :freq"
             ),
             {"freq": frequency},
@@ -1549,6 +1549,23 @@ def sync_worker(self, frequency):
                 sync_counts["sync_skipped"] += 1
             continue
 
+        metadata = doc.get("metadata")
+        if isinstance(metadata, str):
+            try:
+                metadata = json.loads(metadata)
+            except ValueError:
+                metadata = {}
+        if (
+            doc.get("connection_id")
+            and isinstance(metadata, dict)
+            and metadata.get("sync_state") == "paused_reconnect"
+        ):
+            # An S3 or GitHub source whose connection needs reconnecting:
+            # it resumes when the owner reconnects, rather than failing (and
+            # notifying) on every schedule until then.
+            sync_counts["sync_skipped"] += 1
+            continue
+
         source_data = normalize_remote_data(source_type, doc.get("remote_data"))
         if not source_data:
             # No syncable URL/config — skip instead of dispatching a sync
diff --git a/tests/connectors/test_runtime.py b/tests/connectors/test_runtime.py
index cde70b67..3c53e415 100644
--- a/tests/connectors/test_runtime.py
+++ b/tests/connectors/test_runtime.py
@@ -239,6 +239,27 @@ class TestScheduledSync:
         assert counts["sync_dispatched"] == 1
         assert counts["sync_skipped"] == 1
 
+    def test_paused_repository_is_skipped_until_reconnected(self, pg_conn):
+        """A GitHub or S3 source paused for reconnect is not retried (and failed) on every schedule."""
+        from docsgpt import worker
+
+        cid = _connection(pg_conn, provider="github", status="reconnect_needed",
+                          secrets={"credentials": {"access_token": "revoked"}})
+        pg_conn.execute(text(
+            "INSERT INTO sources (user_id, name, type, sync_frequency, connection_id, remote_data, metadata) "
+            "VALUES ('alice', 'acme/api', 'github', 'daily', CAST(:c AS uuid), '{\"repo_url\": \"acme/api\"}', "
+            "'{\"sync_state\": \"paused_reconnect\"}')"
+        ), {"c": cid})
+
+        @contextmanager
+        def _yield():
+            yield pg_conn
+
+        with patch.object(worker, "db_readonly", _yield), patch.object(worker, "sync") as sync:
+            counts = worker.sync_worker(MagicMock(), "daily")
+        sync.assert_not_called()
+        assert counts["sync_skipped"] == 1
+
     def test_paused_connection_is_not_synced(self, pg_conn):
         from docsgpt import worker