homogene 0.1.0

Turn common AI data-processing tasks into simple pandas DataFrame operations.

Homogene

Homogene is a Python library that helps data scientists apply LLM instructions to every row of a pd.DataFrame for data processing.

Running an LLM over a pd.DataFrame column is easy for 10 rows. At 10,000 rows, you're managing threads, catching per-row errors, and checkpointing progress. Homogene handles all of that so you focus on the instruction, not the infrastructure.

Processors

Homogene currently ships with one processor: Generator, which applies a plain-text instruction to every row of a pd.DataFrame independently and in parallel. More processors are coming soon.

Homogene uses LiteLLM to route model calls. Please refer to their documentation for the full list of supported models.

Install

pip install homogene

Quick start

import os
import pandas as pd
from homogene import Generator

os.environ["OPENAI_API_KEY"] = "your-api-key"

df = pd.read_csv("tickets.csv")

generator = Generator(
    df=df,
    columns=["subject", "body"],
    instruction="Identify the main topic of this support ticket in three words or less.",
    model="gpt-5.5",
)

df["topic"] = generator.run()

For a full walkthrough see the Generator example notebook, covering:

• Model configuration
• Freeform generation
• Structured output
• Prompt inspection
• Subset runs
• Checkpointing
• Error handling