Volcengine Agent Development Kit (VeADK) provides observability capabilities to record the critical paths and intermediate states of an agent during execution. It can also report trace data to Application Performance Management (APM) via an exporter, enabling developers to perform debugging, performance analysis, behavior tracing, and other related tasks. This topic describes how to monitor the trace data reported by VeADK through APM.

Background

VeADK is an end-to-end architecture for agents that is launched by Volcano Engine. It provides developers with a full-process development framework for agent build, cloud deployment, and evaluation and optimization. Compared with existing agent development frameworks, VeADK is deeply integrated with the Volcano Engine product ecosystem. It helps developers to build enterprise-grade AI agent applications more efficiently. For more details, refer to the official VeADK documentation.

Prerequisites

VeADK has been installed. For detailed installation steps, refer to the VeADK installation guide.

You have obtained the AK and SK of your Volcano Engine account. To obtain the AK and SK information, refer to How to obtain API access keys.

Procedure

Step 1: Configure environment variables

Configure your Volcano Engine AK and SK as environment variables for your application based on your actual needs. These credentials are used to set the API Key or other authentication information for large model calls.

Step 2: Report observability data

Build your agent by following the instructions in Quick start guide. After the agent is built, configure a cloud exporter to report observability data to APM. Sample code is provided below:

Step 3: Monitor agent tracing

Once VeADK reports observability data as expected, you can view the relevant tracing information for your agent in APM.

Log in to the APM console, and select the project and region in the top navigation bar.

In the left navigation bar, select Service Monitoring > Service List to go to the service list page.

Filter the service associated with the target agent by service name, click the service name to access the service details page, and view the monitoring metrics for the agent on the Service overview tab. Such metrics include QPS of large model calls, number of large model calls, and token usage.

On the Trace analysis tab, click the target trace ID to go to the details page of a single trace. View the agent instrumentation data in the list view.

Observability Platform