NumpyAI

A Natural Language Interface for NumPy powered by LLMs. Empowering mindful data analysis using Generative AI.

About NumpyAI

NumpyAI enables seamless interaction with NumPy using natural language queries, making numerical computing more intuitive and efficient.

Key Features:

• Writes NumPy code for you based on your natural language queries.
• Know what data-analysis steps to apply on your data using numpyai.Diagnosis.
• Talk to multiple arrays using numpyai.NumpyAISession.
• Checks the validity of the generated code.
• Unit tests the code before returning the final-output.
• Full transparency, know what code was executed by the LLM using the verbose=True flag.
• Supports frameworks like sklearn and matplotlib for basic tasks.
• Interactive debugging and re-tries.
• NumPy compatible.

Installation

pip install numpyai

Installation from Source

Clone the project then:

cd numpyai/
pip install -r requirements.txt
pip install .

Setup

Windows

set GOOGLE_API_KEY=...

Linux

export GOOGLE_API_KEY=...

Usage Example

Single Array

import numpyai as npi
import numpy as np

# Ensure GOOGLE_API_KEY environment variable is set.

# Create an array instance
data = [[1, 2, 3, 4, 5, np.nan], [np.nan, 3, 5, 3.1415, 2, 2]]
arr = npi.array(data)

# Query NumPyAI with natural language
print(arr.chat("Compute the height and width of the image using NumPy."))  # Expected output: (2, 6)

Multiple Arrays

import numpyai as npi
import numpy as np

arr1 = np.array([[1, 2, 3], [4, 5, 6]])
arr2 = np.random.random((2, 3))

sess = npi.NumpyAISession([arr1, arr2])
imputed_array = sess.chat("Impute the first array with the mean of the second array.")

Diagnosis

import numpyai as npi
import numpy as np

arr1 = np.array([[1, 2, 3], [4, 5, 6]])
arr2 = np.random.random((2, 3))

sess = npi.NumpyAISession([arr1, arr2])
diag = npi.Diagnosis(sess)
print(diag.steps(task="Tell me the exact and pithy steps to analyse and select which ML model to use for this data. There should be no more than 7 steps"))

Supported LLM Vendors

• Google Gemini
• OpenAI
• Anthropic

Contributing Guidelines

• Apply the black formatter.
• The code should be well documented and be rendered in the docs.
• For testing, add it in the examples/all_functionality.ipynb notebook.
• Ensure backward compatibility.

Thank you and looking forward to seeing you contribute to NumpyAI :) !