Does Uncoil read Semantic Kernel's OpenTelemetry output?

Uncoil accepts OTLP over HTTP with JSON. Many OpenTelemetry exporters send protobuf by default, so check your exporter's protocol before relying on it.

Can I trace planners and multi-step agents?

Yes. Wrap the planner or agent call, and add a step to each plugin function it can call.

Does it work with async kernels?

Yes. Concurrent invocations are kept in separate traces.

Integrations / Semantic Kernel

Trace Semantic Kernel agents

To trace Semantic Kernel with Uncoil in Python, wrap the invocation with trace_agent and open a step around the kernel function or agent call. Plugins you mark show up as steps in order.

Read the guide Start free

Your code

Semantic Kernel

Uncoil

One run.

01One trace per kernel invocation
02A step for each function or plugin you mark
03Inputs, outputs and timing
04Errors in order

Setup

5 steps.

01
Install
The Python flavor of Semantic Kernel and the Uncoil SDK.
bash
```
pip install uncoil-sdk semantic-kernel
```
02
Set your project key
Create a project in Uncoil, copy its API key and put it in the environment where your code runs. The SDK reads it from there, so nothing is hard-coded.
bash
```
export UNCOIL_API_KEY="uncoil_sk_live_..."
```

Wrap the invocation

Semantic Kernel is async. trace_agent and span both work in async functions.

python

import uncoil

@uncoil.trace_agent(name="sk-assistant")
async def ask(question: str) -> str:
    with uncoil.span("kernel-invoke", type="agent", input=question) as step:
        result = await kernel.invoke(plugin_name="chat", function_name="answer", question=question)
        step.set_output(str(result))
        return str(result)

Show plugin calls

A kernel function decorator is the best place for a step, because every call through it is then recorded wherever it comes from.

python

from semantic_kernel.functions import kernel_function

class OrdersPlugin:
    @kernel_function(description="Look up an order by id")
    async def lookup(self, order_id: str) -> str:
        with uncoil.span("orders.lookup", type="tool", input=order_id) as step:
            status = await orders_api.status(order_id)
            step.set_output(status)
            return status

05
Run it once and look
Run your code one time, then open the Tracing page of your project. Your kernel call shows up as a run with its steps in order, plus the tokens, time and any error for each call.

Good to know

Python only for now

The Uncoil SDKs are available for Python and TypeScript. For Semantic Kernel in .NET or Java, you can send traces to the Uncoil API directly.

Good to know

Model calls

Semantic Kernel calls models through its own connectors, so those requests are not wrapped automatically. Mark the function that makes them and you get its timing and output.

Full details in the documentation. All integrations are listed on the integrations page.

Does Uncoil read Semantic Kernel's OpenTelemetry output?: Uncoil accepts OTLP over HTTP with JSON. Many OpenTelemetry exporters send protobuf by default, so check your exporter's protocol before relying on it.
Can I trace planners and multi-step agents?: Yes. Wrap the planner or agent call, and add a step to each plugin function it can call.
Does it work with async kernels?: Yes. Concurrent invocations are kept in separate traces.

Send your first trace.

Free to start.

Start free Compare tools