Structured Output Tracking for Instructor with Opik - Opik

Instructor is a Python library for working with structured outputs for LLMs built on top of Pydantic. It provides a simple way to manage schema validations, retries and streaming responses.

In this guide, we will showcase how to integrate Opik with Instructor so that all the Instructor calls are logged as traces in Opik.

Account Setup

Comet provides a hosted version of the Opik platform, simply create an account and grab your API Key.

You can also run the Opik platform locally, see the installation guide for more information.

Getting Started

Installation

First, ensure you have both opik and instructor installed:

bash

pip install opik instructor

Configuring Opik

Configure the Opik Python SDK for your deployment type. See the Python SDK Configuration guide for detailed instructions on:

CLI configuration: opik configure
Code configuration: opik.configure()
Self-hosted vs Cloud vs Enterprise setup
Configuration files and environment variables

Configuring Instructor

In order to use Instructor, you will need to configure your LLM provider API keys. For this example, we'll use OpenAI, Anthropic, and Gemini. You can find or create your API keys in these pages:

You can set them as environment variables:

bash

export OPENAI_API_KEY="YOUR_API_KEY"
export ANTHROPIC_API_KEY="YOUR_API_KEY"
export GOOGLE_API_KEY="YOUR_API_KEY"

Or set them programmatically:

python

import os
import getpass

if "OPENAI_API_KEY" not in os.environ:
    os.environ["OPENAI_API_KEY"] = getpass.getpass("Enter your OpenAI API key: ")

if "ANTHROPIC_API_KEY" not in os.environ:
    os.environ["ANTHROPIC_API_KEY"] = getpass.getpass("Enter your Anthropic API key: ")

if "GOOGLE_API_KEY" not in os.environ:
    os.environ["GOOGLE_API_KEY"] = getpass.getpass("Enter your Google API key: ")

Using Opik with Instructor library

In order to log traces from Instructor into Opik, we are going to patch the instructor library. This will log each LLM call to the Opik platform.

For all the integrations, we will first add tracking to the LLM client and then pass it to the Instructor library:

python

from opik.integrations.openai import track_openai
import instructor
from pydantic import BaseModel
from openai import OpenAI

# We will first create the OpenAI client and add the `track_openai`
# method to log data to Opik
openai_client = track_openai(OpenAI())

# Patch the OpenAI client for Instructor
client = instructor.from_openai(openai_client)

# Define your desired output structure
class UserInfo(BaseModel):
    name: str
    age: int

user_info = client.chat.completions.create(
    model="gpt-4o-mini",
    response_model=UserInfo,
    messages=[{"role": "user", "content": "John Doe is 30 years old."}],
)

print(user_info)

Thanks to the track_openai method, all the calls made to OpenAI will be logged to the Opik platform. This approach also works well if you are also using the opik.track decorator as it will automatically log the LLM call made with Instructor to the relevant trace.

Integrating with other LLM providers

The instructor library supports many LLM providers beyond OpenAI, including: Anthropic, AWS Bedrock, Gemini, etc. Opik supports the majority of these providers as well.

Here are the code snippets needed for the integration with different providers:

Anthropic

python

from opik.integrations.anthropic import track_anthropic
import instructor
from anthropic import Anthropic

# Add Opik tracking
anthropic_client = track_anthropic(Anthropic())

# Patch the Anthropic client for Instructor
client = instructor.from_anthropic(
    anthropic_client, mode=instructor.Mode.ANTHROPIC_JSON
)

user_info = client.chat.completions.create(
    model="claude-3-5-sonnet-20241022",
    response_model=UserInfo,
    messages=[{"role": "user", "content": "John Doe is 30 years old."}],
    max_tokens=1000,
)

print(user_info)

Gemini

python

from opik.integrations.genai import track_genai
import instructor
from google import genai

# Add Opik tracking
gemini_client = track_genai(genai.Client())

# Patch the GenAI client for Instructor
client = instructor.from_genai(
    gemini_client, mode=instructor.Mode.GENAI_STRUCTURED_OUTPUTS
)

user_info = client.chat.completions.create(
    model="gemini-2.0-flash-001",
    response_model=UserInfo,
    messages=[{"role": "user", "content": "John Doe is 30 years old."}],
)

print(user_info)

You can read more about how to use the Instructor library in their documentation.