Quickstart: Tracing

Overview

Tracing is a powerful tool for understanding the behavior of your LLM application. Phoenix has best-in-class tracing, regardless of what framework you use, and has first-class instrumentation for a variety of frameworks (LlamaIndex, LangChain, DSPy), SDKs (OpenAI, Bedrock, Mistral, Vertex), and Languages (Python, Javascript). You can also manually instrument your application using the OpenTelemetry SDK.

This example will walk you through how to use Phoenix to trace OpenAI requests.

Install & Launch Phoenix

Let's start by installing Phoenix. You have a few options for how to do this:

Phoenix Cloud

The easiest way to use Phoenix is by accessing a free persistent instance provided on our site. Sign up for an Arize Phoenix account at https://app.phoenix.arize.com/login

Once you're there, grab your API key from the Keys option on the left bar:

Self-host

If you'd rather run Phoenix locally, you can instead use one of our self-hosting options. For more detail on each of these, see Self-hosting

Using Terminal

Install the Phoenix package:

pip install arize-phoenix

Launch the Phoenix client:

phoenix serve

This will expose the Phoenix UI and REST API on localhost:6006 and exposes the gRPC endpoint for spans on localhost:4317

Using Docker

Phoenix server images are available via Docker Hub and can be used via docker compose or if you simply want a long-running phoenix instance to share with your team.

docker pull arizephoenix/phoenix:latest

Launch the phoenix docker image using:

docker run -p 6006:6006 -p 4317:4317 arizephoenix/phoenix:latest

This will expose the Phoenix UI and REST API on localhost:6006 and exposes the gRPC endpoint for spans on localhost:4317

Run in Notebook

As a final option, you can run a temporary version of Phoenix directly in your notebook.

Install Phoenix using:

pip install arize-phoenix

Within your notebook, launch Phoenix using:

import phoenix as px
px.launch_app()

By default, notebook instances do not have persistent storage, so your traces will disappear after the notebook is closed. See Persistence or use one of the other deployment options to retain traces.

Connect your application

To collect traces from your application, you must configure an OpenTelemetry TracerProvider to send traces to Phoenix. The register utility from the phoenix.otel module streamlines this process.

Phoenix Cloud

If arize-phoenix is not installed in your python environment, you can use arize-phoenix-otel to quickly connect to your phoenix instance.

pip install arize-phoenix-otel

Set your Phoenix endpoint and API key.

import os

# Add Phoenix API Key for tracing
PHOENIX_API_KEY = "ADD YOUR API KEY"
os.environ["PHOENIX_CLIENT_HEADERS"] = f"api_key={PHOENIX_API_KEY}"
os.environ["PHOENIX_COLLECTOR_ENDPOINT"] = "https://app.phoenix.arize.com"

Self-host

If arize-phoenix is not installed in your python environment, you can use arize-phoenix-otel to quickly connect to your phoenix instance.

pip install arize-phoenix-otel

Set your Phoenix endpoint:

import os

os.environ["PHOENIX_COLLECTOR_ENDPOINT"] = "https://app.phoenix.arize.com"

You do not have to use phoenix.otel to connect to your phoenix instance, you can use OpenTelemetry itself to initialize your OTEL connection. See Using OTEL Python Directly

See Setup Tracing: Python for more details on configuration and setup

Notebook

Phoenix's default register call below will automatically connect to a local notebook instance of Phoenix.

Instrument your application

Now we need to indicate which methods and attributes we want to trace. Phoenix has a number of built-in tracers for popular frameworks, and provides tools to manually instrument your application if needed. See here for a list of integrations

Here we're using OpenAI, so we'll install the built-in OpenAI instrumentor we provide. Based on your needs, install the relevant instrumentors:

pip install -q openinference-instrumentation-openai openai

Initialize auto-instrumentation before your application code:

from phoenix.otel import register

# configure the Phoenix tracer
tracer_provider = register(
  project_name="my-llm-app", # Default is 'default'
  auto_instrument=True # Auto-instrument your app based on installed dependencies
)

Use OpenAI as normal

From here we can use OpenAI as normal. All of our requests will be traced and reported to Phoenix automatically.

# Add OpenAI API Key
import os

os.environ["OPENAI_API_KEY"] = "ADD YOUR OPENAI API KEY"

import openai

client = openai.OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Write a haiku."}],
)
print(response.choices[0].message.content)

View your Traces in Phoenix

You should now see traces in Phoenix!

Next Steps:

• View more details on configuring your tracing

• Run evaluations on traces

• Test changes to you prompts, models, and application via experiments

• Explore other hosting options for Phoenix