Conversational-Shopping-Agent-Project

Project made by Artur Stopa, João Soares and Ricardo Pereira.

Conversational agent built with ML models, such as Clip, Blip and GPT. Read the paper here.

Dialog State Manager Graph

Instalation/Setup

Step 0 - Create env File

Create a .env file with the variables that are in the example. (Given in the labs)

API_USER=''
API_PASSWORD=''
OPENAI_API_KEY=''

Note: If you don't have a Open AI API key, you can run the program without GPT. Everything that isn't a greeting or a recognized intent is treated as OOS(out of scope).

Step 1 - Create a conda environment

conda create -n myenv python=3.9 ipykernel numpy scipy scikit-learn pandas tqdm jupyter matplotlib gensim flask flask_cors ipympl -c defaults -c conda-forge

Then you run that env:

conda activate myenv

Then you can check if you have all the required packages.

conda install pip
pip install -r requirements.txt

Note: From what I understand the pip install -r requirements.txt command will fail if any of the packages listed in the requirements.txt fail to install. Running each line with pip install may be a workaround.

cat requirements.txt | xargs -n 1 pip install

Step 2 - Installing PyTorch+HuggingFace+Spacy

Check lab0 for instructions: https://wiki.novasearch.org/wiki/lab_setup

Step 3 - Import the trained-models folder(for the dialog system)

Get the folder from here: https://drive.google.com/file/d/1VGHkzIzag0bXj4wdD5DD1CTjYU4Fs54D/view

Step 4 - Run the program

python app.py

jpssoares / Conversational-Shopping-Agent-Project