Skip to main content

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

Project description

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"}
)

Project details


Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

openpipe-4.50.0.tar.gz (99.0 kB view details)

Uploaded Source

Built Distribution

openpipe-4.50.0-py3-none-any.whl (440.9 kB view details)

Uploaded Python 3

File details

Details for the file openpipe-4.50.0.tar.gz.

File metadata

  • Download URL: openpipe-4.50.0.tar.gz
  • Upload date:
  • Size: 99.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: poetry/1.7.1 CPython/3.11.0 Darwin/24.3.0

File hashes

Hashes for openpipe-4.50.0.tar.gz
Algorithm Hash digest
SHA256 a2b1bf7a30a8d4c2cf45b85c749839ea9811e36f9d03916df8ffa343d9193a0e
MD5 a1958344bc7157b202c7897dd4f15c96
BLAKE2b-256 ec0b5ac4afd2253e058463fe46b44ebdf9cf153af343b457f13e9e592943c16d

See more details on using hashes here.

File details

Details for the file openpipe-4.50.0-py3-none-any.whl.

File metadata

  • Download URL: openpipe-4.50.0-py3-none-any.whl
  • Upload date:
  • Size: 440.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: poetry/1.7.1 CPython/3.11.0 Darwin/24.3.0

File hashes

Hashes for openpipe-4.50.0-py3-none-any.whl
Algorithm Hash digest
SHA256 2071c3edbba3e08ceb977ad8c12d407f4da86c0c3815447fa33674d918276e5e
MD5 8a4f6511071f82156da3c8f7ddcf8bb9
BLAKE2b-256 923904870a3157d4ad6e8b1671f584da3e064750ccd64aa08339c6fc6dbd3a1c

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page