openpipe 4.50.0

OpenPipe Python SDK: Fine-Tuning, Inference, and Metrics for Production Apps

OpenPipe Python Client

This client allows you automatically report your OpenAI calls to OpenPipe.

Installation

pip install openpipe

Usage

1. Create a project at https://app.openpipe.ai
2. Find your project's API key at https://app.openpipe.ai/settings
3. Configure the OpenPipe client as shown below.

from openpipe import OpenAI
import os

client = OpenAI(
    # defaults to os.environ.get("OPENAI_API_KEY")
    api_key="My API Key",
    openpipe={
        # Set the OpenPipe API key you got in step (2) above.
        # If you have the `OPENPIPE_API_KEY` environment variable set we'll read from it by default
        "api_key": "My OpenPipe API Key",
    }
)

You can now use your new OpenAI client, which functions identically to the generic OpenAI client while also reporting calls to your OpenPipe instance.

Special Features

Metadata Tagging

OpenPipe follows OpenAI’s concept of metadata tagging for requests. This is very useful for grouping a certain set of completions together. These tags will help you find all the input/output pairs associated with a certain prompt and fine-tune a model to replace it. Here's how you can use the tagging feature:

completion = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "system", "content": "count to 10"}],
    metadata={"prompt_id": "counting"},
    store=True,
)

Should I Wait to Enable Logging?

We recommend keeping request logging turned on from the beginning. If you change your prompt you can just set a new prompt_id tag so you can select just the latest version when you're ready to create a dataset.

Usage with langchain

Assuming you have created a project and have the openpipe key.

from openpipe.langchain_llm import ChatOpenAI
from langchain.prompts import ChatPromptTemplate
from langchain.schema.runnable import RunnableSequence

prompt = ChatPromptTemplate.from_messages(
    [
        (
            "system",
            "Classify user query into positive, negative or neutral.",
        ),
        ("human", "{query}"),
    ]
)
llm = ChatOpenAI(model="gpt-4o")
    .with_tags(chain_name="classify", any_key="some")

# To provide the openpipe key explicitly
# llm = ChatOpenAI(model="gpt-4o", openpipe_kwargs={"api_key": "My OpenPipe API Key"})\
#     .with_tags(chain_name="classify", any_key="some")

chain: RunnableSequence = prompt | llm
res = chain.invoke(
    {"query": "this is good"}
)