helpit 1.0.0

Ask OpenAI-compatible models about a Python object with runtime metadata and optional help() snippets.

helpit

Python's help() is great… until it isn’t.

helpit lets you ask questions about help() output and get answers tailored to the exact object in front of you.

---

Install

pip install helpit

---

Quickstart

from openai import OpenAI
from helpit import helpit, set_default_client

set_default_client(OpenAI())  

import torch

x = torch.randn(1, 32, 1)
helpit(
    x.squeeze,
    "How do I remove the last dimension and keep the leading dim?",
)

---

Examples

import pandas as pd
from helpit import helpit

df = pd.DataFrame({"city": [None, "ZRH", None], "sales": [3, 10, 2]})

helpit(df, "Group by city, keep NaNs, and sum sales—what's the right dropna setting?")

from helpit import helpit

def stream():
    for i in range(1_000_000):
        yield i

it = stream()
helpit(it, "How do I safely consume only the first 10 items without exhausting the iterator?")

from sklearn.ensemble import RandomForestClassifier
from helpit import helpit

rf = RandomForestClassifier(
    n_estimators=300,
    max_depth=None,
    min_samples_split=2,
    min_samples_leaf=1,
    max_features="sqrt",
    random_state=0,
)

helpit(rf, "Which hyperparameters matter most for overfitting here? Show how to adjust them.")

---

When docs matter 📚 (grounded answers)

helpit can attach the most relevant help() snippets when add_documentation=True, keeping answers grounded without dumping full docs.

from transformers import pipeline
from helpit import helpit

pipe = pipeline("sentiment-analysis")

helpit(pipe, "How does this pipeline work?", add_documentation=True)

---

Run locally 🏠 (fast + private)

helpit talks to the OpenAI Responses API. Any compatible local server works.

from openai import OpenAI
from helpit import helpit

client = OpenAI(
    base_url="http://localhost:11434/v1",
    api_key="ollama",   # any token works
)

helpit(
    len,
    "How does len behave on nested lists?",
    model="llama3.2",
    openai_client=client,
)

---

Tests

python3 -m unittest discover -s tests -v