vipin wiki

Evaluate Agents

11 min read

Evaluate Agents

Source

Classification

Summary

Evaluating Agents with Langfuse In this cookbook, we will learn how to monitor the internal steps (traces) of the OpenAI agent SDK and evaluate its performance using Langfuse. This guide covers online and offline evaluation metrics used by teams to bring agents to production fast and reliably. To learn more about evaluation strategies, check out this blog po…

What This Teaches

  • How to structure agent workflows, tool use, memory, evaluation, or multi-agent coordination.
  • How to turn model behavior into measurable evaluation cases and improvement loops.

Implementation Use Cases

  • Use as a concrete implementation reference when building OpenAI API systems in this category.
  • Compare against current official API docs before copying model names, SDK calls, or parameters into production code.
  • Preserve this page as a mirrored source; prefer synthesis pages for personal recommendations or project-specific decisions.

Mirrored Content

Evaluating Agents with Langfuse

In this cookbook, we will learn how to monitor the internal steps (traces) of the OpenAI agent SDK and evaluate its performance using Langfuse.

This guide covers online and offline evaluation metrics used by teams to bring agents to production fast and reliably. To learn more about evaluation strategies, check out this blog post.

Why AI agent Evaluation is important:

  • Debugging issues when tasks fail or produce suboptimal results
  • Monitoring costs and performance in real-time
  • Improving reliability and safety through continuous feedback

Step 0: Install the Required Libraries

Below we install the openai-agents library (the OpenAI Agents SDK), the pydantic-ai[logfire] OpenTelemetry instrumentation, langfuse and the Hugging Face datasets library

%pip install openai-agents nest_asyncio "pydantic-ai[logfire]" langfuse datasets

Step 1: Instrument Your Agent

In this notebook, we will use Langfuse to trace, debug and evaluate our agent.

Note: If you are using LlamaIndex or LangGraph, you can find documentation on instrumenting them here and here.

import os
import base64
 
# Get keys for your project from the project settings page: https://cloud.langfuse.com
os.environ["LANGFUSE_PUBLIC_KEY"] = "pk-lf-..."
os.environ["LANGFUSE_SECRET_KEY"] = "sk-lf-..."
os.environ["LANGFUSE_HOST"] = "https://cloud.langfuse.com" # 🇪🇺 EU region
# os.environ["LANGFUSE_HOST"] = "https://us.cloud.langfuse.com" # 🇺🇸 US region
 
# Build Basic Auth header.
LANGFUSE_AUTH = base64.b64encode(
    f"{os.environ.get('LANGFUSE_PUBLIC_KEY')}:{os.environ.get('LANGFUSE_SECRET_KEY')}".encode()
).decode()
 
# Configure OpenTelemetry endpoint & headers
os.environ["OTEL_EXPORTER_OTLP_ENDPOINT"] = os.environ.get("LANGFUSE_HOST") + "/api/public/otel"
os.environ["OTEL_EXPORTER_OTLP_HEADERS"] = f"Authorization=Basic {LANGFUSE_AUTH}"
 
# Your openai key
os.environ["OPENAI_API_KEY"] = "sk-proj-..."

With the environment variables set, we can now initialize the Langfuse client. get_client() initializes the Langfuse client using the credentials provided in the environment variables.

from langfuse import get_client
 
langfuse = get_client()
 
# Verify connection
if langfuse.auth_check():
    print("Langfuse client is authenticated and ready!")
else:
    print("Authentication failed. Please check your credentials and host.")

Pydantic Logfire offers an instrumentation for the OpenAi Agent SDK. We use this to send traces to the Langfuse OpenTelemetry Backend.

import nest_asyncio
nest_asyncio.apply()
import logfire
 
# Configure logfire instrumentation.
logfire.configure(
    service_name='my_agent_service',
 
    send_to_logfire=False,
)
# This method automatically patches the OpenAI Agents SDK to send logs via OTLP to Langfuse.
logfire.instrument_openai_agents()

Step 2: Test Your Instrumentation

Here is a simple Q&A agent. We run it to confirm that the instrumentation is working correctly. If everything is set up correctly, you will see logs/spans in your observability dashboard.

import asyncio
from agents import Agent, Runner
 
async def main():
    agent = Agent(
        name="Assistant",
        instructions="You are a senior software engineer",
    )
 
    result = await Runner.run(agent, "Tell me why it is important to evaluate AI agents.")
    print(result.final_output)
 
loop = asyncio.get_running_loop()
await loop.create_task(main())
 
langfuse.flush()

Check your Langfuse Traces Dashboard to confirm that the spans and logs have been recorded.

Example trace in Langfuse:

Example trace in Langfuse

Link to the trace

Step 3: Observe and Evaluate a More Complex Agent

Now that you have confirmed your instrumentation works, let’s try a more complex query so we can see how advanced metrics (token usage, latency, costs, etc.) are tracked.

import asyncio
from agents import Agent, Runner, function_tool
 
# Example function tool.
@function_tool
def get_weather(city: str) -> str:
    return f"The weather in {city} is sunny."
 
agent = Agent(
    name="Hello world",
    instructions="You are a helpful agent.",
    tools=[get_weather],
)
 
async def main():
    result = await Runner.run(agent, input="What's the weather in Berlin?")
    print(result.final_output)
 
loop = asyncio.get_running_loop()
await loop.create_task(main())

Trace Structure

Langfuse records a trace that contains spans, which represent each step of your agent’s logic. Here, the trace contains the overall agent run and sub-spans for:

  • The tool call (get_weather)
  • The LLM calls (Responses API with ‘gpt-4o’)

You can inspect these to see precisely where time is spent, how many tokens are used, and so on:

Trace tree in Langfuse

Link to the trace

Online Evaluation

Online Evaluation refers to evaluating the agent in a live, real-world environment, i.e. during actual usage in production. This involves monitoring the agent’s performance on real user interactions and analyzing outcomes continuously.

We have written down a guide on different evaluation techniques here.

Common Metrics to Track in Production

  1. Costs — The instrumentation captures token usage, which you can transform into approximate costs by assigning a price per token.
  2. Latency — Observe the time it takes to complete each step, or the entire run.
  3. User Feedback — Users can provide direct feedback (thumbs up/down) to help refine or correct the agent.
  4. LLM-as-a-Judge — Use a separate LLM to evaluate your agent’s output in near real-time (e.g., checking for toxicity or correctness).

Below, we show examples of these metrics.

1. Costs

Below is a screenshot showing usage for gpt-4o calls. This is useful to see costly steps and optimize your agent.

Costs

Link to the trace

2. Latency

We can also see how long it took to complete each step. In the example below, the entire run took 7 seconds, which you can break down by step. This helps you identify bottlenecks and optimize your agent.

Latency

Link to the trace

3. Additional Attributes

Langfuse allows you to pass additional attributes to your spans. These can include user_id, tags, session_id, and custom metadata. Enriching traces with these details is important for analysis, debugging, and monitoring of your application’s behavior across different users or sessions.

In this example, we pass a user_id, session_id and trace_tags to Langfuse.

input_query = "Why is AI agent evaluation important?"
 
with langfuse.start_as_current_span(
    name="OpenAI-Agent-Trace",
    ) as span:
 
    # Run your application here
    async def main(input_query):
            agent = Agent(
                name = "Assistant",
                instructions = "You are a helpful assistant.",
            )
 
            result = await Runner.run(agent, input_query)
            print(result.final_output)
            return result
 
    result = await main(input_query)
 
    # Pass additional attributes to the span
    span.update_trace(
        input=input_query,
        output=result,
        user_id="user_123",
        session_id="my-agent-session",
        tags=["staging", "demo", "OpenAI Agent SDK"],
        metadata={"email": "user@langfuse.com"},
        version="1.0.0"
        )
 
# Flush events in short-lived applications
langfuse.flush()

Example trace in Langfuse

4. User Feedback

If your agent is embedded into a user interface, you can record direct user feedback (like a thumbs-up/down in a chat UI). Below is an example using IPython.display for simple feedback mechanism.

In the code snippet below, when a user sends a chat message, we capture the OpenTelemetry trace ID. If the user likes/dislikes the last answer, we attach a score to the trace.

from agents import Agent, Runner, WebSearchTool
from opentelemetry.trace import format_trace_id
import ipywidgets as widgets
from IPython.display import display
from langfuse import get_client
 
langfuse = get_client()
 
# Define your agent with the web search tool
agent = Agent(
    name="WebSearchAgent",
    instructions="You are an agent that can search the web.",
    tools=[WebSearchTool()]
)
 
def on_feedback(button):
    if button.icon == "thumbs-up":
      langfuse.create_score(
            value=1,
            name="user-feedback",
            comment="The user gave this response a thumbs up",
            trace_id=trace_id
        )
    elif button.icon == "thumbs-down":
      langfuse.create_score(
            value=0,
            name="user-feedback",
            comment="The user gave this response a thumbs down",
            trace_id=trace_id
        )
    print("Scored the trace in Langfuse")
 
user_input = input("Enter your question: ")
 
# Run agent
with langfuse.start_as_current_span(
    name="OpenAI-Agent-Trace",
    ) as span:
 
    # Run your application here
    result = Runner.run_sync(agent, user_input)
    print(result.final_output)
 
    result = await main(user_input)
    trace_id = langfuse.get_current_trace_id()
 
    span.update_trace(
        input=user_input,
        output=result.final_output,
    )
 
# Get feedback
print("How did you like the agent response?")
 
thumbs_up = widgets.Button(description="👍", icon="thumbs-up")
thumbs_down = widgets.Button(description="👎", icon="thumbs-down")
 
thumbs_up.on_click(on_feedback)
thumbs_down.on_click(on_feedback)
 
display(widgets.HBox([thumbs_up, thumbs_down]))
 
# Flush events in short-lived applications
langfuse.flush()

User feedback is then captured in Langfuse:

User feedback is being captured in Langfuse

5. LLM-as-a-Judge

LLM-as-a-Judge is another way to automatically evaluate your agent’s output. You can set up a separate LLM call to gauge the output’s correctness, toxicity, style, or any other criteria you care about.

Workflow:

  1. You define an Evaluation Template, e.g., “Check if the text is toxic.”
  2. You set a model that is used as judge-model; in this case gpt-4o-mini.
  3. Each time your agent generates output, you pass that output to your “judge” LLM with the template.
  4. The judge LLM responds with a rating or label that you log to your observability tool.

Example from Langfuse:

LLM-as-a-Judge Evaluation Template LLM-as-a-Judge Evaluator

# Example: Checking if the agent’s output is toxic or not.
from agents import Agent, Runner, WebSearchTool
 
# Define your agent with the web search tool
agent = Agent(
    name="WebSearchAgent",
    instructions="You are an agent that can search the web.",
    tools=[WebSearchTool()]
)
 
input_query = "Is eating carrots good for the eyes?"
 
# Run agent
with langfuse.start_as_current_span(name="OpenAI-Agent-Trace") as span:
    # Run your agent with a query
    result = Runner.run_sync(agent, input_query)
 
    # Add input and output values to parent trace
    span.update_trace(
        input=input_query,
        output=result.final_output,
    )

You can see that the answer of this example is judged as “not toxic”.

LLM-as-a-Judge Evaluation Score

6. Observability Metrics Overview

All of these metrics can be visualized together in dashboards. This enables you to quickly see how your agent performs across many sessions and helps you to track quality metrics over time.

Observability metrics overview

Offline Evaluation

Online evaluation is essential for live feedback, but you also need offline evaluation—systematic checks before or during development. This helps maintain quality and reliability before rolling changes into production.

Dataset Evaluation

In offline evaluation, you typically:

  1. Have a benchmark dataset (with prompt and expected output pairs)
  2. Run your agent on that dataset
  3. Compare outputs to the expected results or use an additional scoring mechanism

Below, we demonstrate this approach with the search-dataset, which contains questions that can be answered via the web search tool and expected answers.

import pandas as pd
from datasets import load_dataset
 
# Fetch search-dataset from Hugging Face
dataset = load_dataset("junzhang1207/search-dataset", split = "train")
df = pd.DataFrame(dataset)
print("First few rows of search-dataset:")
print(df.head())

Next, we create a dataset entity in Langfuse to track the runs. Then, we add each item from the dataset to the system.

from langfuse import get_client
langfuse = get_client()
 
langfuse_dataset_name = "search-dataset_huggingface_openai-agent"
 
# Create a dataset in Langfuse
langfuse.create_dataset(
    name=langfuse_dataset_name,
    description="search-dataset uploaded from Huggingface",
    metadata={
        "date": "2025-03-14",
        "type": "benchmark"
    }
)
for idx, row in df.iterrows():
    langfuse.create_dataset_item(
        dataset_name=langfuse_dataset_name,
        input={"text": row["question"]},
        expected_output={"text": row["expected_answer"]}
    )
    if idx >= 49: # For this example, we upload only the first 50 items
        break

Dataset items in Langfuse

Running the Agent on the Dataset

We define a helper function run_openai_agent() that:

  1. Starts a Langfuse span
  2. Runs our agent on the prompt
  3. Records the trace ID in Langfuse

Then, we loop over each dataset item, run the agent, and link the trace to the dataset item. We can also attach a quick evaluation score if desired.

from agents import Agent, Runner, WebSearchTool
from langfuse import get_client
 
langfuse = get_client()
dataset_name = "search-dataset_huggingface_openai-agent"
current_run_name = "qna_model_v3_run_05_20" # Identifies this specific evaluation run
 
agent = Agent(
    name="WebSearchAgent",
    instructions="You are an agent that can search the web.",
    tools=[WebSearchTool(search_context_size= "high")]
)
 
# Assume 'run_openai_agent' is your instrumented application function
def run_openai_agent(question):
    with langfuse.start_as_current_generation(name="qna-llm-call") as generation:
        # Simulate LLM call
        result = Runner.run_sync(agent, question)
 
        # Update the trace with the input and output
        generation.update_trace(
            input= question,
            output=result.final_output,
        )
 
        return result.final_output
 
dataset = langfuse.get_dataset(name=dataset_name) # Fetch your pre-populated dataset
 
for item in dataset.items:
 
    # Use the item.run() context manager
    with item.run(
        run_name=current_run_name,
        run_metadata={"model_provider": "OpenAI", "temperature_setting": 0.7},
        run_description="Evaluation run for Q&A model v3 on May 20th"
    ) as root_span: # root_span is the root span of the new trace for this item and run.
        # All subsequent langfuse operations within this block are part of this trace.
 
        # Call your application logic
        generated_answer = run_openai_agent(question=item.input["text"])
 
        print(item.input)

You can repeat this process with different:

  • Search tools (e.g. different context sized for OpenAI’s WebSearchTool)
  • Models (gpt-4o-mini, o1, etc.)
  • Tools (search vs. no search)

Then compare them side-by-side in Langfuse. In this example, I did run the agent 3 times on the 50 dataset questions. For each run, I used a different setting for the context size of OpenAI’s WebSearchTool. You can see that an increased context size also slightly increased the answer correctness from 0.89 to 0.92. The correct_answer score is created by an LLM-as-a-Judge Evaluator that is set up to judge the correctness of the question based on the sample answer given in the dataset.

Dataset run overview Dataset run comparison

Graph View

Backlinks