Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
Python agentagentic-aiagentic-frameworkagentic-workflowaiautomationclaudedeepseekgenaigptllmlow-codemcpnextjsno-codeopenaipythonskillsworkflow
fix(api): silence opentelemetry.context error reports to Sentry on cloud (#39841)
Bond Zhuc562f78d4f7e
diff --git a/api/extensions/ext_sentry.py b/api/extensions/ext_sentry.pyindex 69d1f1ab07e..1d4c071d5f5 100644--- a/api/extensions/ext_sentry.py+++ b/api/extensions/ext_sentry.py@@ -7,6 +7,7 @@ def init_app(app: DifyApp):import sentry_sdkfrom sentry_sdk.integrations.celery import CeleryIntegrationfrom sentry_sdk.integrations.flask import FlaskIntegration+ from sentry_sdk.integrations.logging import ignore_loggerfrom werkzeug.exceptions import HTTPExceptionfrom graphon.model_runtime.errors.invoke import InvokeRateLimitError@@ -45,3 +46,12 @@ def init_app(app: DifyApp):release=f"dify-{dify_config.project.version}-{dify_config.COMMIT_SHA}",before_send=before_send,)++ if dify_config.BILLING_ENABLED:+ # Cloud only. `opentelemetry.context.detach()` catches its own failures and reports+ # them through `logger.exception`, so a double-detach surfaces as an error event+ # rather than as a raised exception. Under gevent this can fire about once per+ # workflow run, which is enough volume to crowd real errors out of the issue stream.+ # Self-hosted deployments rarely see enough of it to be worth silencing, and the+ # records still reach the normal logging handlers either way.+ ignore_logger("opentelemetry.context")