> ## Documentation Index
> Fetch the complete documentation index at: https://docs.keywordsai.co/llms.txt
> Use this file to discover all available pages before exploring further.

# Langfuse

> Send Langfuse traces to Keywords AI with OTEL instrumentation.

<Note>Instrument before importing Langfuse so the HTTP client is patched.</Note>

## Overview

Use the Keywords AI Langfuse instrumentor to redirect Langfuse traces to Keywords AI without changing your Langfuse usage. You can keep using `@observe` and the Langfuse SDK as usual.

## Quickstart

### Step 1: Get a Keywords AI API key

Create an API key in the [Keywords AI dashboard](https://platform.keywordsai.co/platform/api/api-keys).

<Frame className="rounded-md">
  <img src="https://keywordsai-static.s3.us-east-1.amazonaws.com/docs/documentation/admin/kai_create_api_key_v0.png" alt="Create a Keywords AI API key" />
</Frame>

### Step 2: Install packages

<CodeGroup>
  ```bash Python theme={"system"}
  pip install langfuse keywordsai-instrumentation-langfuse
  ```
</CodeGroup>

### Step 3: Set environment variables

Create a `.env` or export environment variables:

```bash .env theme={"system"}
KEYWORDSAI_API_KEY=your-keywordsai-api-key
KEYWORDSAI_BASE_URL=https://api.keywordsai.co/api

LANGFUSE_PUBLIC_KEY=your-langfuse-public-key
LANGFUSE_SECRET_KEY=your-langfuse-secret-key
LANGFUSE_HOST=https://cloud.langfuse.com
```

## Examples

Below are three common Langfuse patterns using the Keywords AI instrumentor.

#### Basic decorator

Instrument the Langfuse client and use `@observe`.

<CodeGroup>
  ```python Python theme={"system"}
  import dotenv
  dotenv.load_dotenv(".env", override=True)

  import os
  from keywordsai_instrumentation_langfuse import LangfuseInstrumentor

  # Instrument BEFORE importing Langfuse
  instrumentor = LangfuseInstrumentor()
  instrumentor.instrument(
      api_key=os.environ["KEYWORDSAI_API_KEY"],
      endpoint=os.environ["KEYWORDSAI_BASE_URL"] + "/v1/traces/ingest",
  )

  from langfuse import Langfuse, observe

  langfuse = Langfuse(
      public_key="pk-lf-...",
      secret_key="sk-lf-...",
      host="https://cloud.langfuse.com",
  )

  @observe()
  def process_query(query: str):
      return f"Response to: {query}"

  result = process_query("Hello World")
  print(result)

  langfuse.flush()
  ```
</CodeGroup>

<Frame className="rounded-md">
  <img src="https://keywordsai-static.s3.us-east-1.amazonaws.com/docs/Integrations/Langfuse/basic_decorator.png" alt="Langfuse basic decorator tracing in Keywords AI" />
</Frame>

#### Generation tracing

Mark a function as a generation with `@observe(as_type="generation")`.

<CodeGroup>
  ```python Python theme={"system"}
  import os
  from keywordsai_instrumentation_langfuse import LangfuseInstrumentor

  os.environ["KEYWORDSAI_API_KEY"] = "your-api-key"

  instrumentor = LangfuseInstrumentor()
  instrumentor.instrument(
      api_key=os.environ["KEYWORDSAI_API_KEY"],
  )

  from langfuse import Langfuse, observe

  langfuse = Langfuse(
      public_key="pk-lf-...",
      secret_key="sk-lf-...",
  )

  @observe(as_type="generation")
  def generate_response(prompt: str):
      return f"Generated: {prompt}"

  result = generate_response("Write a poem")
  print(result)

  langfuse.flush()
  ```
</CodeGroup>

<Frame className="rounded-md">
  <img src="https://keywordsai-static.s3.us-east-1.amazonaws.com/docs/Integrations/Langfuse/generation-tracing.png" alt="Langfuse generation tracing in Keywords AI" />
</Frame>

#### Nested traces

Create parent-child relationships with nested `@observe` functions.

<CodeGroup>
  ```python Python theme={"system"}
  import os
  from keywordsai_instrumentation_langfuse import LangfuseInstrumentor

  os.environ["KEYWORDSAI_API_KEY"] = "your-api-key"

  instrumentor = LangfuseInstrumentor()
  instrumentor.instrument(
      api_key=os.environ["KEYWORDSAI_API_KEY"],
  )

  from langfuse import Langfuse, observe

  langfuse = Langfuse(
      public_key="pk-lf-...",
      secret_key="sk-lf-...",
  )

  @observe()
  def subtask(name: str):
      return f"Completed: {name}"

  @observe()
  def main_workflow(task: str):
      result1 = subtask("step 1")
      result2 = subtask("step 2")
      return f"Workflow done: {task}"

  result = main_workflow("Process request")
  print(result)

  langfuse.flush()
  ```
</CodeGroup>

<Frame className="rounded-md">
  <img src="https://keywordsai-static.s3.us-east-1.amazonaws.com/docs/Integrations/Langfuse/nested_traces.png" alt="Langfuse nested traces in Keywords AI" />
</Frame>