Hi dudleykong, Thanks for your question! Here’s the current state of observability support in Vapi: Current Status - **Langfuse Integration**: Vapi has native integration with Langfuse, an open-source observability platform for AI applications. This is currently the primary officially supported option. Custom Observability Options While Vapi doesn’t yet offer native support for other providers like LangSmith, you do have a few paths to implement custom observability: 1. Vapi’s Built-in Features - Vapi assistants include an

observabilityPlan

system. - The platform also exposes boards, evals, and scorecard features in the documentation that may be helpful for tracking and evaluating behavior. 2. OpenTelemetry Integration - You can capture logs, metrics, and traces from Vapi’s operations. - Export telemetry data to any platform that supports the OpenTelemetry standard (including LangSmith). 3. Webhook-based Monitoring - Vapi supports webhooks, which can capture call events in real-time. - You could forward these events into your observability platform and build your own dashboards. Recommendation If you want to go beyond the official Langfuse integration, I’d suggest: - Reviewing the latest observability docs in Vapi. - Exploring webhook setups for call event monitoring. - Using OpenTelemetry as a bridge if your preferred observability provider supports it. This way, you can extend observability to LangSmith (or any other platform) while continuing to use Vapi.