Audit Logs Stop Working After Upgrading to LangSmith v0.13.37

Symptom

After upgrading a self-hosted LangSmith instance from v0.13.36 (or earlier) to v0.13.37, audit logs stop being recorded. Previously logged operations (e.g., PAT creation, API key changes) no longer appear. The can_use_audit_logs permission is correctly set to true and no errors are visible in the UI.

Root Cause

In v0.13.37, the audit log write path was changed from the Python SAQ queue to the Go asynq queue. A new environment variable GO_AUDIT_LOG_QUEUE_ENABLED was introduced and defaults to true.

When enabled, audit log entries are sent to the Go platform backend's /internal/enqueue endpoint instead of the Python SAQ queue. If the Go platform backend is not reachable from the Python backend or the Go asynq worker is not processing audit log tasks, entries are silently dropped.

Resolution

Add the following to your Helm values.yaml to revert to the previous (Python SAQ) write path:

backend:
  deployment:
    extraEnv:
      - name: GO_AUDIT_LOG_QUEUE_ENABLED
        value: "false"

Then apply with:

helm upgrade <release-name> langsmith/langsmith -f values.yaml

Audit logs should resume immediately after the backend pods restart.