⚰️(summary) cleaned tasks failure handling

Code in task failure was assuming that the failure signal would be called on retry which is not the case.
✨(summary) report exception type in failure analytics
2026-07-27 12:19:10 +00:00 · 2026-07-24 12:07:44 +02:00 · 2026-07-24 10:14:59 +02:00 · 2026-07-24 10:02:39 +02:00 · 2026-07-24 09:58:58 +02:00 · 2026-07-22 14:23:10 +02:00
5 changed files with 62 additions and 70 deletions
@@ -8,9 +8,15 @@ and this project adheres to

 ## [Unreleased]

+### Added
+
+- ✨(summary) report exception type in failure analytics
+
 ## Fixed

 - 🐛(transcription) fix silent bug in speaker assignment
+- 🐛(summary) extend tasks auto retry logic
+- 🐛(summary) properly detect when failure webhook should be sent

 ### Changed

@@ -18,6 +24,7 @@ and this project adheres to
 - ⬆️(frontend) upgrade i18next from 26.3.1 to 26.3.2
 - ⬆️(frontend) upgrade posthog-js from 1.391.2 to 1.395.0
 - ⬆️(frontend) upgrade @tanstack/react-query from 5.101.0 to 5.101.1
+- ⬆️(frontend) upgrade livekit-client from 2.19.2 to 2.20.0

 ## [1.24.0] - 2026-07-21

@@ -29,7 +29,7 @@
        "i18next-browser-languagedetector": "8.2.1",
        "i18next-parser": "9.4.0",
        "i18next-resources-to-backend": "1.2.1",
-        "livekit-client": "2.19.2",
+        "livekit-client": "2.20.0",
        "posthog-js": "1.395.0",
        "react": "18.3.1",
        "react-aria": "3.50.0",
@@ -1028,9 +1028,9 @@
      "license": "Apache-2.0"
    },
    "node_modules/@livekit/protocol": {
-      "version": "1.45.8",
-      "resolved": "https://registry.npmjs.org/@livekit/protocol/-/protocol-1.45.8.tgz",
-      "integrity": "sha512-Q+l57E7w/xxOBFVWzdX5rkAZO7ffyF+rlDzNUYq2SU114+5aTyCq+PK4unaEVDNd4952Af7wteKr3sOgasGuaA==",
+      "version": "1.46.6",
+      "resolved": "https://registry.npmjs.org/@livekit/protocol/-/protocol-1.46.6.tgz",
+      "integrity": "sha512-upzlHP1vi/kZ/QqALZTFskQ0ifqc2f15RKucHYOsIHJsaXvEYanG75mAb7o+Yomfs4XhQ4BaRsdY+TFHXpaqrg==",
      "license": "Apache-2.0",
      "dependencies": {
        "@bufbuild/protobuf": "^1.10.0"
@@ -8125,13 +8125,13 @@
      "license": "MIT"
    },
    "node_modules/livekit-client": {
-      "version": "2.19.2",
-      "resolved": "https://registry.npmjs.org/livekit-client/-/livekit-client-2.19.2.tgz",
-      "integrity": "sha512-Kvk07QYDWRAbmYNLRll04ZIuxMQobW/oLPYnmR1kCy8GGHpU0gqyHf704Rz+29zfy8IJZRjKqeVbzGSKn9sumw==",
+      "version": "2.20.0",
+      "resolved": "https://registry.npmjs.org/livekit-client/-/livekit-client-2.20.0.tgz",
+      "integrity": "sha512-RIJcpvBmOmwz3jTj3rmdY6Dzr55HrhcaJjMgY+HSmoEM+yIRyA40m7r8UKv0hnZWM3z/AYhP1q8C8ciz5UWFKQ==",
      "license": "Apache-2.0",
      "dependencies": {
        "@livekit/mutex": "1.1.1",
-        "@livekit/protocol": "1.45.8",
+        "@livekit/protocol": "1.46.6",
        "events": "^3.3.0",
        "jose": "^6.1.0",
        "loglevel": "^1.9.2",
@@ -36,7 +36,7 @@
    "i18next-browser-languagedetector": "8.2.1",
    "i18next-parser": "9.4.0",
    "i18next-resources-to-backend": "1.2.1",
-    "livekit-client": "2.19.2",
+    "livekit-client": "2.20.0",
    "posthog-js": "1.395.0",
    "react": "18.3.1",
    "react-aria": "3.50.0",
@@ -202,13 +202,16 @@ class MetadataManager:
            },
        )

-    def capture(self, task_id, event_name):
+    def capture(self, task_id, event_name, extra_properties=None):
        """Capture analytics event with task metadata and clean up."""
        if self._is_disabled or not self.has_task_id(task_id):
            return

        metadata = self._get_metadata(task_id)

+        if extra_properties:
+            metadata = {**metadata, **extra_properties}
+
        if "start_time" in metadata:
            metadata["execution_time"] = round(time.time() - metadata["start_time"], 2)
            del metadata["start_time"]
@@ -431,7 +431,7 @@ def _should_auto_create_summary(payload: TranscribeTaskJob) -> bool:
@celery.task(
    max_retries=3,
    queue=settings.call_webhook_queue_v2,
-    autoretry_for=[exceptions.HTTPError],
+    autoretry_for=[exceptions.RequestException],
 )
 def call_webhook_v2_task(
    payload: dict,
@@ -445,7 +445,9 @@ def call_webhook_v2_task(

@celery.task(
    bind=True,
-    autoretry_for=[exceptions.HTTPError],
+    autoretry_for=[
+        exceptions.RequestException,
+    ],
    max_retries=settings.celery_max_retries,
    queue=settings.transcribe_queue_v2,
 )
@@ -575,12 +577,6 @@ def task_retry_handler_transcript(request=None, reason=None, einfo=None, **kwarg
    metadata_manager.retry(request.id)


-@signals.task_failure.connect(sender=process_audio_transcribe_v2_task)
-def task_failure_handler_transcript(task_id, exception=None, **kwargs):
-    """Signal handler called when task execution fails permanently."""
-    metadata_manager.capture(task_id, settings.posthog_transcript_failure)
-
-
@signals.task_failure.connect(sender=process_audio_transcribe_v2_task)
 def handle_transcribe_v2_failed(
    sender,
@@ -594,30 +590,26 @@ def handle_transcribe_v2_failed(
 ):
    """Handle the failure of transcribe_v2_task.

-    This function is triggered when the transcribe_v2_task fails.
-    It sends a webhook failure payload to notify the client of the failure.
+    Tracks the failure event in analytics and sends a failure webhook to the client.
    """
-    n_retries_left = sender.max_retries - sender.request.retries - 1
-    if n_retries_left > 0:
-        logger.info(
-            "Transcribe task %s failed, %s retries left.",
-            task_id,
-            n_retries_left,
-        )
-    else:
-        logger.warn(
-            "Transcribe task %s failed, no more retries left, sending failure webhook.",
-            task_id,
-        )
-        call_webhook_v2_task.apply_async(
-            args=[
-                TranscribeWebhookFailurePayload(
-                    job_id=task_id,
-                    error_code="unknown_error",
-                ).model_dump(),
-                args[0]["tenant_id"],
-            ]
-        )
+    logger.error(
+        "Transcribe task %s failed, no more retries left, sending failure webhook.",
+        task_id,
+    )
+    metadata_manager.capture(
+        task_id,
+        settings.posthog_transcript_failure,
+        {"exception_type": type(exception).__name__},
+    )
+    call_webhook_v2_task.apply_async(
+        args=[
+            TranscribeWebhookFailurePayload(
+                job_id=task_id,
+                error_code="unknown_error",
+            ).model_dump(),
+            args[0]["tenant_id"],
+        ]
+    )


@celery.task(
@@ -681,12 +673,6 @@ def task_retry_handler_summary(request=None, reason=None, einfo=None, **kwargs):
    metadata_manager.retry(request.id)


-@signals.task_failure.connect(sender=summarize_v2_task)
-def task_failure_handler_summary(task_id, exception=None, **kwargs):
-    """Signal handler called when task execution fails permanently."""
-    metadata_manager.capture(task_id, settings.posthog_summary_failure)
-
-
@signals.task_failure.connect(sender=summarize_v2_task)
 def handle_summarize_v2_failed(
    sender,
@@ -700,27 +686,23 @@ def handle_summarize_v2_failed(
 ):
    """Handle the failure of summarize_v2_task.

-    This function is triggered when the summarize_v2_task fails.
-    It sends a webhook failure payload to notify the client of the failure.
+    Tracks the failure event in analytics and sends a failure webhook to the client.
    """
-    n_retries_left = sender.max_retries - sender.request.retries - 1
-    if n_retries_left > 0:
-        logger.info(
-            "Summary task %s failed, %s retries left.",
-            task_id,
-            n_retries_left,
-        )
-    else:
-        logger.warn(
-            "Summary task %s failed, no more retries left, sending failure webhook.",
-            task_id,
-        )
-        call_webhook_v2_task.apply_async(
-            args=[
-                SummarizeWebhookFailurePayload(
-                    job_id=task_id,
-                    error_code="unknown_error",
-                ).model_dump(),
-                args[0]["tenant_id"],
-            ]
-        )
+    logger.warn(
+        "Summary task %s failed, no more retries left, sending failure webhook.",
+        task_id,
+    )
+    metadata_manager.capture(
+        task_id,
+        settings.posthog_summary_failure,
+        {"exception_type": type(exception).__name__},
+    )
+    call_webhook_v2_task.apply_async(
+        args=[
+            SummarizeWebhookFailurePayload(
+                job_id=task_id,
+                error_code="unknown_error",
+            ).model_dump(),
+            args[0]["tenant_id"],
+        ]
+    )
Author	SHA1	Message	Date
Florent Chehab	a5b79afde1	⚰️(summary) cleaned tasks failure handling Code in task failure was assuming that the failure signal would be called on retry which is not the case.	2026-07-24 12:07:44 +02:00
Florent Chehab	052d3c1b22	✨(summary) report exception type in failure analytics We know capture also the exception type that was raised in failure analytics to provide more insights about what happened.	2026-07-24 10:14:59 +02:00
Florent Chehab	dff5aa9575	🐛(summary) properly detect if task will be retried In previous version, webhook may not be called in case of failure, because the task wouldn't be actually retried. We know check the exception raised against the auto_retry for config. Analytics capture also happens in case of definitive failure.	2026-07-24 10:02:39 +02:00
Florent Chehab	ded93bf24f	🐛(summary) retry on RequestException instead of HTTPError Request may fail for other reasons than HTTPError (ConnectionError for instance). This change switches to the more generalized error for the retry logic on call webhook & transcribe audio	2026-07-24 09:58:58 +02:00
leo	5ba1885411	✅(transcription) fix broken speaker assignment tests Fix broken speaker assignement tests following #1522.	2026-07-22 14:23:10 +02:00
snyk-bot	273af221e6	⬆️(frontend) upgrade livekit-client from 2.19.2 to 2.20.0 Snyk has created this PR to upgrade livekit-client from 2.19.2 to 2.20.0. See this package in npm: livekit-client See this project in Snyk: https://app.eu.snyk.io/org/lasuite-dinum-default/project/96ea03d8-8d09-493d-86bf-363f274e129e?utm_source=github&utm_medium=referral&page=upgrade-pr	2026-07-21 19:00:04 +02:00