Quick Start

GPTree

# Setup imports

from IPython.display import Image
import pandas as pd
import numpy as np
from think_reason_learn.gptree import GPTree
from think_reason_learn.core.llms import GoogleChoice, OpenAIChoice
from think_reason_learn.core.llms import XAIChoice, AnthropicChoice

import asyncio

# Sample data: Predict startup founder success

X = pd.DataFrame(
    {
        "founder_info": [
            "Alex is a serial entrepreneur with two successful exits, strong network in Silicon Valley, and expertise in AI.",
            "Jordan graduated top of class from Oxford but has no prior business experience and limited funding.",
            "Taylor has 10 years in finance, secured seed funding quickly, and built a talented team.",
            "Casey started a company right out of high school, faced multiple failures, but persists with innovative ideas.",
            "Morgan is a former Google engineer with patents in machine learning and venture capital backing."
            "Seraphine graduated from UMaT with a first class degree in Minerals Engineering and has a strong network in the mining industry."
        ]
    }
)
y = np.array(["successful", "failed", "successful", "failed", "successful", "successful"])

# Initialize GPTree with LLM choices
tree = GPTree(
    qgen_llmc=[
        GoogleChoice(model="gemini-1.5-flash-latest"),
        OpenAIChoice(model="gpt-5"),
        XAIChoice(model="grok-3-mini"),
    ],
    critic_llmc=[
        XAIChoice(model="grok-3-mini"),
        OpenAIChoice(model="gpt-4o-mini"),
        AnthropicChoice(model="claude-3-5-sonnet-20240620"),
    ],
    qgen_instr_llmc=[
        GoogleChoice(model="gemini-1.5-flash-latest"),
        XAIChoice(model="grok-3-mini"),
    ],
)

# Generate question generation instructions template
qgit = await tree.set_tasks(
    task_description="Predict if a startup founder will be successful or fail based on their background.",
)

# Print the generated instructions template
print(qgit)

# Fit the tree
fitter = tree.fit(X, y, reset=True)

# Build the root node / Could do it in a loop or manually
root = await anext(fitter)

# Visualize the tree
display(Image(tree.view_tree()))

# Get training data with generated features
tree.get_training_data()

# Get all generated questions
tree.get_questions()

# Predict on the training data
# NOTE: ensure the fitter is fully consumed
predictions = await tree.predict(X)
for pred in predictions:
    print(pred)

For more examples and comprehensive use cases, visit the examples directory on our GitHub repository.