Build Tool definitions for OpenAI Function Calling API

The OpenAI function calling API provides a specification for extracting details from text and representing the outputs in a json object. This is very useful in situations where you want to provide the extracted features to a downstream process for analysis.

The purpose of this application is to provide a user interface for composing a tool specification to perform a specific data extraction task.

Test the application here: https://toolbuilder.streamlit.app

Note the option to load example data to demonstrate the application.

Run locally

Clone the project and enter the project directory.

git clone https://github.com/nhoffman/toolbuilder.git
cd toolbuilder

You will need to provide an OpenAI API key, either using an environment variable, or within the app. For example:

export OPENAI_API_KEY="..."

You can also specify an alternative API endpoint using OPENAI_BASE_URL.

The easiest way to launch the app is with uv:

uv run --with-requirements requirements.txt streamlit run toolbuilder.py

Without uv:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
streamlit run toolbuilder.py

The functions for interacting with the OpenAI API are defined in utils.py. You can test the function calling API from the command line like this (requires that OPENAI_API_KEY is defined):

% uv -q run --script utils.py
{'label': 'A', 'gleason_score': '3+4=7', 'biopsy_length': 2.1, 'cancer_length': 0.1, 'diagnosis': 'malignant'}
{'label': 'B', 'gleason_score': '', 'biopsy_length': 0, 'cancer_length': 0, 'diagnosis': 'benign'}
{'label': 'C', 'gleason_score': '', 'biopsy_length': 0, 'cancer_length': 0, 'diagnosis': 'benign'}
{'label': 'D', 'gleason_score': '3+4=7', 'biopsy_length': 2.5, 'cancer_length': 0.8, 'diagnosis': 'malignant'}
{'label': 'E', 'gleason_score': '3+4=7', 'biopsy_length': 2.2, 'cancer_length': 1.4, 'diagnosis': 'malignant'}
{'label': 'F', 'gleason_score': '3+4=7', 'biopsy_length': 2.9, 'cancer_length': 0.3, 'diagnosis': 'malignant'}
{'label': 'G', 'gleason_score': '3+4=7', 'biopsy_length': 2.0, 'cancer_length': 1.2, 'diagnosis': 'malignant'}

Deploy

Deployment to Streamlit community cloud is performed by launching the app locally and using the interactive "deploy" action.