Doc: Test integration of langfuse + instructor with openai / gemini using openai's python SDK

[Mr-Ruben]

In case someone finds it useful.
If It can be improved, even better. 
This is just one way to do it.
You can test this very quickly on https://colab.research.google.com/

instructor https://github.com/instructor-ai/instructor
langfuse https://cloud.langfuse.com/

Test integration of langfuse + instructor with openai / gemini using openai's python SDK

LANGFUSE_SECRET_KEY="sk-..."
LANGFUSE_PUBLIC_KEY="pk-..."
LANGFUSE_HOST="https://cloud.langfuse.com" # different if in US

OPENAI_API_KEY='sk-...'
GEMINI_API_KEY= '...'

# Install
!pip install langfuse  openai  instructor

# Import required modules
# from openai import OpenAI  # Not anymore
import langfuse
import instructor

# This wraps OpenAI to include Langfuse tracking
from langfuse.openai import openai

# Configure Langfuse credentials
openai.langfuse_public_key = LANGFUSE_PUBLIC_KEY
openai.langfuse_secret_key = LANGFUSE_SECRET_KEY
openai.langfuse_host = LANGFUSE_HOST

# Validate Langfuse authentication (not needed, just for testing)
openai.langfuse_auth_check()

True

# now, patch the clients with instructor
#!pip install instructor
import instructor

# Create clients with Langfuse integration patched by Instructor
client_openai_with_langfuse = instructor.patch(
    openai.OpenAI(api_key=OPENAI_API_KEY)
)

# Gemini can use the same python SDK, just need different KEY and HOST
client_gemini_with_langfuse = instructor.patch(
    openai.OpenAI(api_key=GEMINI_API_KEY,
        base_url="https://generativelanguage.googleapis.com/v1beta/openai/"
    )
)

model_openai= "gpt-4o-mini"
model_gemini= "gemini-2.0-flash"

# A simple example of call to llm using instructor's response_model

from pydantic import BaseModel

class Output(BaseModel):
    name: str
    age: int

prompt="John has 3 children, 1 car and is 33 years old."

# send call
response=client_openai_with_langfuse.chat.completions.create(
    model=model_openai,
    messages=[{"role": "user", "content": prompt}],
    response_model=Output,
    tags=["test","openai"], # optional: for langfuse only
)

print(response)

response=client_gemini_with_langfuse.chat.completions.create(
    model=model_gemini,
    messages=[{"role": "user", "content": prompt}],
    response_model=Output,
    tags=["test","gemini"], # optional: for langfuse only
)

print(response)

name='John' age=33
name='John' age=33