llmaix 0.0.6

Add your description here

LLMAIx (v2) Library

The llmaix library contains the core functionality of the LLMAIx framework.

[!CAUTION] The interface of the library is still in development and may change in the future. The library is not yet ready for production use.

Features

• Preprocessing: The library provides tools for extracting text from various file formats, including PDF, DOCX, and TXT. It can apply OCR to images and PDFs, using tesseract, surya-ocr and VLMs via docling.

• Information Extraction: The library provides a wrapper helping you to get a JSON response from an LLM. All OpenAI-API compatible models are supported!

Installation

pip install llmaix

To install dependencies for docling:

pip install llmaix[docling]

Available Dependency groups: surya,docling

To install all dependencies:

pip install llmaix[all]

Usage

CLI

llmaix --help

Python

Preprocessing a PDF file without OCR:

from llmaix import preprocess_file

filename = "tests/testfiles/987462_text.pdf"

extracted_text = preprocess_file(filename)

Preprocessing a PDF file with OCR:

from llmaix import preprocess_file

filename = "tests/testfiles/987462_notext.pdf"

extracted_text = preprocess_file(filename, use_ocr=True, ocr_backend="ocrmypdf")

OCR Backends	Comment
ocrmypdf	Uses tesseract. Needs to be installed on the system first!
surya-ocr	Uses surya-ocr. Runs models via transformers library locally.
doclingvlm	Uses docling to perform OCR using a VLM. Configure the model like for information extraction!

PDF Backends	Comment
pymupdf4llm	Uses pymupdf to extract text as markdown from PDF files.
markitdown	Uses markitdown to extract text as markdown from PDF files.
docling	Uses docling to extract text as markdown from PDF files. Caution: docling itself might apply OCR even if you don't specify it.
ocr_backend	Directly use the text output from the OCR backend. Incompatible with ocrmypdf.

Extracting information from a text:

1. Provide a .env file with your OpenAI API key:

echo "OPENAI_API_KEY=your_openai_api_key" > .env

2. (Optional) To use a custom base url, set the OPENAI_API_BASE environment variable:

echo "OPENAI_API_BASE=https://your_custom_base_url/v1" >> .env

2. (Optional) Configure model in the .env file:

echo "OPENAI_MODEL=gpt-4o-2024-08-06" >> .env

3. Use the extract_info function to extract information from a text. In this example, a pydantic model is used to define the expected output format. The output will be a JSON object.

from llmaix import extract_info
from pydantic import BaseModel

extracted_text = "The KatherLab is a research group at the University of Technology Dresden, lead by Prof. Jakob N. Kather."

class LabInformation(BaseModel):
    name: str
    location: str
    lead: str

extracted_info = extract_info(
    prompt=f"Extract the name, location and lead of the lab from the following text: {extracted_text}",
    llm_model="Llama-4-Maverick-17B-128E-Instruct-FP8",
    pydantic_model=LabInformation,
)

Clone the repository and install the dependencies:

git clone https://github.com/KatherLab/LLMAIx-v2.git
cd LLMAIx-v2
uv sync

Tests

Run the tests using the following command:

uv run pytest

Example to just run test for preprocessing with the ocrmypdf backend:

uv run pytest tests/test_preprocess.py --ocr-backend ocrmypdf