GPT4DFCI-API 🔌

Welcome to the code repository for GPT4DFCI-API, an example of how to call the API of GPT4DFCI, which a allows you to leverage GPT4DFCI programmatically in your application. GPT4DFCI is a private and secure generative AI tool, based on GPT and o1 models that deployed for non-clinical use at Dana-Farber Cancer Institute. See all details about GPT4DFCI in the separate GPT4DFCI code repository.

😈 If you are an advanced user, and just want the GPT4DFCI-API code straightaway please see demo.py.

🤓 All other GPT4DFCI-API users, please keep reading. This repository is organized in the following sections:

• Prerequisites
• Run the GPT4DFCI API demo
• License
• Contact

⚙️ Prerequisites

Poetry is a tool for dependency management and packaging in Python.

If you are already familiar with Poetry, or are comfortable setting up your own environment with other tools, you can probably skip this.

Installing Poetry

https://python-poetry.org/docs/#installation

For macOS this would be the suggested way:

• Install brew:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

• Install poetry using: brew install poetry

Installing GPT4DFCI API dependencies

After installing Poetry, navigate to the directory you are interested in and run:

poetry install --no-root

This should create a virtual environment with the dependencies specified in pyproject.toml installed.

To activate the virtual environment, run:

poetry shell

or, alternatively, if you don't want to use Poetry's sub-shell pattern, you can run: source .venv/bin/activate

🟢 Run the GPT4DFCI API demo

Set up environment variables

In a file called .env, fill in the following values:

AZURE_OPENAI_ENDPOINT=https://... AZURE_OPENAI_ENTRA_SCOPE=XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX

AZURE_OPENAI_ENDPOINT should be the GPT4DFCI API endpoint, without the /openai suffix.

AZURE_OPENAI_ENTRA_SCOPE should be the scope required for the bearer token, the UUID client_id.

Authenticate via Azure CLI

Prerequisites: Azure CLI

For macOS this would be

brew install azure-cli If you are using DFCI AI Studio, you should use %sh curl -sL https://aka.ms/InstallAzureCLIDeb | sudo bash Then, make sure you are logged in via the Azure CLI to an account that has been granted access to the GPT4DFCI API.

az login --allow-no-subscriptions

Close the window after logging in and get back to the shell.

VPN

Turn on the VPN or make sure you are running from an on-prem resource.

Run the demo

export $(cat .env | xargs) && python demo.py

Shutting down

When finished working, shut the current virtual environment with

deactivate

Resuming work

When you are read to resume working on this, come back to this folder and resume working with

source .venv/bin/activate export $(cat .env | xargs) && python demo.py

🎫 License

The GNU GPL v2 version of GPT4DFCI is made available via Open Source licensing. The user is free to use, modify, and distribute under the terms of the GNU General Public License version 2.

Commercial license options are available also, and include additional features.

📧 Contact

Questions? Comments? Suggestions? Get in touch!

aios@dfci.harvard.edu