# Observability for LlamaIndex Workflows

This cookbook demonstrates how to use [Langfuse](/content/site-root.html) to gain real-time observability for your [LlamaIndex Workflows](https://docs.llamaindex.ai/en/stable/module_guides/workflow/).

> **What are LlamaIndex Workflows?** [LlamaIndex Workflows](https://docs.llamaindex.ai/en/stable/module_guides/workflow/) is a flexible, event-driven framework designed to build robust AI agents. In LlamaIndex, workflows are created by chaining together multiple steps—each defined and validated using the `@step` decorator. Every step processes specific event types, allowing you to orchestrate complex processes such as AI agent collaboration, RAG flows, data extraction, and more.

> **What is Langfuse?** [Langfuse](/content/site-root.html) is the open source AI engineering platform. It helps teams to collaboratively manage prompts, trace applications, debug problems, and evaluate their LLM system in production.

## Get Started

We'll walk through a simple example of using LlamaIndex Workflows and integrating it with Langfuse.

### Step 1: Install Dependencies

```
%pip install langfuse openai llama-index-workflows llama-index-core llama-index-llms-openai openinference-instrumentation-llama_index llama-index-instrumentation
```

### Step 2: Set Up Environment Variables

Configure your Langfuse API keys. You can get them by signing up for [Langfuse Cloud](https://cloud.langfuse.com/) or [self-hosting Langfuse](/content/self-hosting/index.html).

```
import os

# Get keys for your project from the project settings page: https://cloud.langfuse.com
os.environ.setdefault("LANGFUSE_PUBLIC_KEY", "pk-lf-...")
os.environ.setdefault("LANGFUSE_SECRET_KEY", "sk-lf-...")
os.environ.setdefault("LANGFUSE_BASE_URL", "https://cloud.langfuse.com") # 🇪🇺 EU region
# Other Langfuse data regions include 🇺🇸 US: https://us.cloud.langfuse.com, 🇯🇵 Japan: https://jp.cloud.langfuse.com and ⚕️ HIPAA: https://hipaa.cloud.langfuse.com

# Your openai key
os.environ.setdefault("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.")
```

### Step 3: Initialize LlamaIndex Instrumentation

Now, we initialize the [OpenInference LlamaIndex instrumentation](https://docs.arize.com/phoenix/tracing/integrations-tracing/llamaindex). This third-party instrumentation automatically captures LlamaIndex operations and exports OpenTelemetry (OTel) spans to Langfuse.

```
from openinference.instrumentation.llama_index import LlamaIndexInstrumentor

# Initialize LlamaIndex instrumentation
LlamaIndexInstrumentor().instrument()
```

### Step 4: Create a Simple LlamaIndex Workflows Application

In LlamaIndex Workflows, you build event-driven AI agents by defining steps with the `@step` decorator. Each step processes an event and, if appropriate, emits new events. In this example, we create a simple workflow with two steps: one that pre-processes an incoming event and another that generates a reply.

```
from llama_index.core.llms import ChatMessage
from llama_index.llms.openai import OpenAI
from typing import Annotated

from workflows import Workflow, step
from workflows.events import StartEvent, StopEvent
from workflows.resource import Resource

def get_llm(**kwargs):
    return OpenAI(model="gpt-4.1-mini")

class MyWorkflow(Workflow):
    @step
    async def step1(
        self, ev: StartEvent, llm: Annotated[OpenAI, Resource(get_llm)]
    ) -> StopEvent:
        msg = ChatMessage(role="user", content=ev.get("input"))
        response = await llm.achat([msg])
        return StopEvent(result=response.message.content)

w = MyWorkflow()
```

```
response = await w.run(input="Hello, what is Langfuse?")
print(response)
```

### Step 5: View Traces in Langfuse

After running your workflow, log in to [Langfuse](https://cloud.langfuse.com/) to explore the generated traces. You will see logs for each workflow step along with metrics such as token counts, latencies, and execution paths.

_[Public example trace in Langfuse](https://cloud.langfuse.com/project/cloramnkj0002jz088vzn1ja4/traces/e0987dce85c8c49602030599b84e77e8?timestamp=2025-07-01T09%3A42%3A54.701Z&display=details)_

ℹ️

_**Note:** To add additional trace attributes like tags or metadata or use LlamaIndex Workflows together with other Langfuse features please refer to [this guide](/content/integrations/frameworks/llamaindex/index.html)._

## Interoperability with the Python SDK

You can use this integration together with the Langfuse [SDKs](/content/docs/observability/sdk/overview/index.html) to add additional attributes to the observation.

DecoratorContext Manager

The [`@observe()` decorator](/content/docs/observability/sdk/instrumentation#custom-instrumentation/index.html) provides a convenient way to automatically wrap your instrumented code and add additional attributes to the observation.

```
from langfuse import observe, propagate_attributes, get_client

langfuse = get_client()

@observe()
def my_llm_pipeline(input):
    # Add additional attributes (user_id, session_id, metadata, version, tags) to all spans created within this execution scope
    with propagate_attributes(
        user_id="user_123",
        session_id="session_abc",
        tags=["agent", "my-observation"],
        metadata={"email": "user@langfuse.com"},
        version="1.0.0"
    ):

# YOUR APPLICATION CODE HERE
        result = call_llm(input)

return result

# Run the function
my_llm_pipeline("Hi")
```

Learn more about using the Decorator in the [Langfuse SDK instrumentation docs](/content/docs/observability/sdk/instrumentation#custom-instrumentation/index.html).

The [Context Manager](/content/docs/observability/sdk/instrumentation#custom-instrumentation/index.html) allows you to wrap your instrumented code using context managers (with `with` statements), which allows you to add additional attributes to the observation.

```
from langfuse import get_client, propagate_attributes

langfuse = get_client()

with langfuse.start_as_current_observation(
    as_type="span",
    name="my-observation",
    trace_context={"trace_id": "abcdef1234567890abcdef1234567890"},  # Must be 32 hex chars
) as observation:

# Add additional attributes (user_id, session_id, metadata, version, tags)
    # to all observations created within this execution scope
    with propagate_attributes(
        user_id="user_123",
        session_id="session_abc",
        metadata={"experiment": "variant_a", "env": "prod"},
        version="1.0",
    ):
        # YOUR APPLICATION CODE HERE
        result = call_llm("some input")

# Flush events in short-lived applications
langfuse.flush()
```

Learn more about using the Context Manager in the [Langfuse SDK instrumentation docs](/content/docs/observability/sdk/instrumentation#custom-instrumentation/index.html).

## Troubleshooting

No observations appearing

First, enable [debug mode](/content/docs/observability/sdk/advanced-features#logging--debugging/index.html) in the Python SDK:

```
export LANGFUSE_DEBUG="True"
```

Then run your application and check the debug logs:

- **OTel observations appear in the logs:**Your application is instrumented correctly but observations are not reaching Langfuse. To resolve this:
1. Call [`langfuse.flush()`](/content/docs/observability/sdk/instrumentation#client-lifecycle--flushing/index.html) at the end of your application to ensure all observations are exported.
2. Verify that you are using the correct API keys and base URL.
- **No OTel spans in the logs:** Your application is not instrumented correctly. Make sure the instrumentation runs before your application code.

Unwanted observations in Langfuse

The Langfuse SDK is based on OpenTelemetry. Other libraries in your application may emit OTel spans that are not relevant to you. These still count toward your [billable units](/content/docs/administration/billable-units/index.html), so you should filter them out. See [Unwanted spans in Langfuse](/content/faq/all/unwanted-http-database-spans/index.html) for details.

Missing attributes

Some attributes may be stored in the metadata object of the observation rather than being mapped to the Langfuse data model. If a mapping or integration does not work as expected, please [raise an issue on GitHub](/content/issues/index.html).

## Next Steps

Once you have instrumented your code, you can manage, evaluate and debug your application:

[**Manage prompts in Langfuse**](/content/docs/prompts/get-started/index.html) [**Add evaluation scores**](/content/docs/evaluation/features/evaluation-methods/custom-scores/index.html) [**Run LLM-as-a-judge Evaluators**](/content/docs/scores/model-based-evals/index.html) [**Create datasets**](/content/docs/datasets/overview/index.html) [**Create custom dashboards**](/content/docs/analytics/custom-dashboards/index.html) [**Test queries in the Playground**](/content/docs/playground/index.html)

## References

- [LlamaIndex Workflows Documentation](https://docs.llamaindex.ai/en/stable/module_guides/workflow/)