Semantic Kernel Agent Telemetry

Implementation evidence

Semantic Kernel Agent Telemetry: from boundary to verified row

Use Semantic Kernel Agent Telemetry at a controlled application boundary, keep the event contract small, and verify a known outcome before building aggregate views.

1
Choose the outcome
Semantic Kernel agent monitoring
2
Define the contract
operation_id, workflow, plugin_name, function_name, agent_name, model_alias, and release
3
Instrument the boundary
Wrap Kernel.InvokeAsync or the owning agent boundary and emit one terminal application outcome after completion.
4
Verify the evidence
Exercise a known fixture, then inspect agent_run_completed for one correctly typed terminal row.

Before you start

Prerequisites and boundaries

Semantic Kernel configured in a trusted .NET service
A shared Telemetry Log API wrapper with a short timeout and server-side key
Stable plugin, function, agent, model, workflow, and release names

Delivery setup

Install and initialize server-side

Use one shared HttpClient with a server-side key, a short timeout, and a small allowlisted Log API wrapper. Keep ingestion credentials out of browser bundles, client-visible environment variables, source control, logs, and exception messages.

1Prepare one reusable server-side delivery client with bounded network behavior.
2Add the outcome event at the success, failure, retry, or timeout boundary.
3Send controlled fixtures and inspect the stored rows before enabling an alert.

Snippet

Start with one structured event

Add this shape where the workflow completes, fails, or retries. Then build the dashboard from real fields.

semantic-kernel

Semantic Kernel Agent Telemetry event

csharp

var startedAt = Stopwatch.GetTimestamp();
var status = "success";
string? errorType = null;

try
{
    return await kernel.InvokeAsync(
        "SupportPlugin",
        "ResolveCase",
        arguments,
        cancellationToken
    );
}
catch (Exception error)
{
    status = "failed";
    errorType = ClassifyAgentError(error);
    throw;
}
finally
{
    await telemetry.LogAsync("agent_run_completed", new
    {
        operation_id = operationId,
        workflow = "support_resolution",
        plugin_name = "SupportPlugin",
        function_name = "ResolveCase",
        status,
        error_type = errorType,
        duration_ms = Stopwatch.GetElapsedTime(startedAt).TotalMilliseconds,
        release = appRelease,
    }, cancellationToken);
}

Event contract

operation_id, workflow, plugin_name, function_name, agent_name, model_alias, and release

status, duration_ms, tool_call_count, retry_count, human_handoff, and error_type

input_tokens, output_tokens, estimated_cost_usd, accepted, and environment when approved

Implementation checkpoints

Checkpoint 1

Wrap Kernel.InvokeAsync or the owning agent boundary and emit one terminal application outcome after completion.

Checkpoint 2

Semantic Kernel emits OpenTelemetry-compatible logs, metrics, and traces. Keep detailed spans in that backend instead of copying all span attributes into Telemetry.

Checkpoint 3

Do not send prompts, chat history, function arguments, function results, tool payloads, model content, credentials, or unrestricted exception messages.

Verification

Prove the event arrived

Run this after exercising known success and failure cases. Replace the fallback table name if your final event contract differs from the snippet.

semantic-kernel-verification

Semantic Kernel Agent Telemetry verification query

sql

SELECT *
FROM agent_run_completed
ORDER BY timestamp_utc DESC
LIMIT 20;

Confirm one terminal row per logical outcome, with the expected status, identifiers, units, and UTC time.

Inspect the inferred schema and verify that retries do not change field types or generate a new logical event ID.

Search the stored fields for credentials, raw payloads, prompts, private content, and unbounded error messages.

Exercise a provider timeout, ingestion rejection, and process shutdown before treating the dashboard as complete.

Implementation references

Review the event contract, data-safety guidance, and upstream primary documentation before enabling a new production path.

OpenTelemetry GenAI outcome events Nested tool-call SQL LLM time to first token SQL Semantic Kernel observability Semantic Kernel InvokeAsync API

Production boundary

Keep the outcome event small and recoverable

This pattern provides

A bounded, SQL-ready outcome beside the upstream workflow.
Stable fields for dashboards, alerts, and cross-event correlation.
A fixture-driven path for validating success, failure, retry, and timeout behavior.

This pattern does not provide

An OTLP exporter, automatic collection pipeline, or replacement for detailed traces and diagnostic logs.
Exactly-once delivery merely because the payload contains an event ID.
Permission to collect raw provider payloads, user content, credentials, or regulated data.

Delivery and idempotency Batching and shutdown Troubleshoot ingestion

Event schema starting points

Review the row grain, emit boundary, required types, privacy classes, example payload, and validation checklist before adapting a query or snippet to production.

AI operations11 fields