Seamless Integration of Python, Pandas, and LLM

These days, it’s straightforward to make use of totally different massive language fashions (LLMs) through the online interface or the general public API. However can we seamlessly combine LLM into the information evaluation course of and use the mannequin immediately from Python or Jupyter Pocket book? Certainly, we will, and on this article, I’ll present three alternative ways to do it. As traditional, all elements used within the article can be found at no cost.

Let’s get into it!

1. Pandas AI

The primary Python library I’m going to check is Pandas AI. It permits us to ask questions on our Pandas dataframe in pure language. As a toy instance, I created a small dataframe with all EU nations and their populations:

import pandas as pd
df = pd.DataFrame({
"Nation": ['Austria', 'Belgium', 'Bulgaria', 'Croatia', 'Cyprus', 'Czech Republic', 'Denmark', 'Estonia', 'Finland',
'France', 'Germany', 'Greece', 'Hungary', 'Iceland', 'Ireland', 'Italy', 'Latvia', 'Liechtenstein', 'Lithuania',
'Luxembourg', 'Malta', 'Monaco', 'Montenegro', 'Netherlands', 'Norway', 'Poland', 'Portugal', 'Romania', 'Serbia',
'Slovakia', 'Slovenia', 'Spain', 'Sweden', 'Switzerland'],
"Inhabitants": [8_205000, 10_403000, 7_148785, 4_491000, 1_102677, 10_476000, 5_484000, 1_291170, 5_244000,
64_768389, 82_369000, 11_000000, 9_930000, 308910, 4_622917, 58_145000, 2_217969, 35000, 3_565000,
497538, 403000, 32965, 666730, 16_645000, 4_907000, 38_500000, 10_676000, 21_959278, 7_344847,
5_455000, 2_007000, 46_505963, 9_045000, 7_581000]
})
df.to_csv('knowledge.csv', index=False)

Earlier than utilizing Pandas AI, let’s create the LLM occasion:

from pandasai.llm.local_llm import LocalLLM
from pandasai.llm import OpenAI
# Native LLM
pandas_llm = LocalLLM(api_base="http://localhost:8000/v1")
OR
# OpenAI
pandas_llm = OpenAI(api_token="...")

Right here, we’ve two decisions. These readers who’ve an OpenAI API key can use the OpenAI class. It’s going to present higher and quicker outcomes, however this API is clearly not free. Another choice is to make use of a LocalLLM occasion, which makes use of the OpenAI-compatible server…