Skip to content

Home

Jump to bottom
Michela Tjan edited this page Apr 10, 2025 · 5 revisions

Welcome to the ACUNAO wiki!

Description

An AI assistant as a domain expert of the lab’s documents that has the ability to answer users’ queries accurately. The user must be able to trust the answers given by the AI assistant and give input in order for the AI assistant to continuously learn how to better respond to users’ queries.

Due to the rapid growth of AI, especially since the launch of ChatGPT, there has been a drastic increase in AI products and services. These products and services have mostly been developed to appeal to the general public. There is a lack of AI solution to support interdisciplinary research labs whose documents could accumulate so quickly, which leads to misplaced or forgotten information. The nature of these documents also poses as a problem since they are generally complex and confidential. ACUNAO aims to bridge this gap.

Wiki Purpose

The purpose is this wiki is to set guidelines for GitHub practices and document issues, solutions, and ideas. Examples of good wikis can be found here: https://github.com/MyHoneyBadger/awesome-github-wiki?tab=readme-ov-file

Testing Document

As of July 23, 2024, testing has been done using the ips-revised-handbook-2022-2023.pdf on questions found in this Excel sheet: link to Excel sheet

Table of Contents

Download LLM

Please follow the following instructions prior to installing the AI Assistant.

  1. Navigate to the following link https://huggingface.co/bartowski/Phi-3.5-mini-instruct-GGUF/resolve/main/Phi-3.5-mini-instruct-Q8_0.gguf?download=true.

Installation

Follow the proper instructions based on your package manager, environment management system, and operating system. If you use Conda as your environment management system,follow the instructions found in Conda-environment. If you use MacOS without Conda, follow the instructions found in MacOS no conda.

To set up ACUNAO AI locally, follow these steps in your terminal or VS Code terminal:

Conda-environment

  1. Clone the repository or Download the repository 
git clone https://github.com/luquelab/ACUNAO-AI.git
  1. Navigate to project directory
cd ACUNAO-AI

or

cd ACUNAO-AI-main

Note: if you are new to the terminal, PowerShell or CommandPrompt, run ls to see the current directory and cd to navigate to the ACUNAO-AI folder.

  1. Add the LLM
    Move the downloaded LLM to the ACUNAO-AI/utils/data/ folder, replacing the existing pointer file: Phi-3.5-mini-instruct-Q8_0.gguf.

  2. Install dependencies

conda env create --name venv --file=environment.yml

conda activate venv

sh ./post_install.sh
  1. Run the application

Run the Streamlit prototype:

python run.py

Note: During first run, you will be prompted to add your email for marketing subscription from Streamlit. You have the option to add your email and press enter, or hit the enter key to skip adding your email to the marketing subscription.

MacOS No Conda

  1. Clone the repository or Download the repository 
git clone https://github.com/luquelab/ACUNAO-AI.git
  1. Navigate to project directory
cd ACUNAO-AI

or

cd ACUNAO-AI-main

Note: if you are new to the terminal, PowerShell or CommandPrompt, run ls to see the current directory and cd to navigate to the ACUNAO-AI folder.

  1. Add the LLM
    Move the downloaded LLM to the ACUNAO-AI/utils/data/ folder.

  2. Create a virtual environment
    Note: In MacOS, make sure that Xcode is installed. To do this, run this command in your terminal:

xcode-select –-install

python3 -m venv venv
  1. Activate the virtual environment
source venv/bin/activate
  1. Install dependencies

This project requires Tesseract to be installed on your system. You can install Tesseract using Homebrew with the following command:

brew install tesseract

Then:

pip install -r requirements.txt

And then:

sh ./post_install.sh
  1. Run the application

Run the Streamlit prototype:

python run.py

Note: During first run, you will be prompted to add your email for marketing subscription from Streamlit. You have the option to add your email and press enter, or hit the enter key to skip adding your email to the marketing subscription.

Project History

This is an evolving repository Started: 2024-06-03

End: Ongoing

Maintainers

@LuqueLab @TjanMichela

Contributors

This project exists thanks to all the people who contribute.
@LuqueLab @TjanMichela

License

This project is licensed under MIT license.

The syntax of markdown files (.md) is CommonMark unless specified otherwise (https://commonmark.org/help/)

Clone this wiki locally