This repository contains the code for the master's thesis "Intent Detection Module for a Conversational Assistant".
The module requires Python 3.8, pip3 and pipenv installed on your machine.
Clone the repository and install dependencies.
cd intent_recognition
pipenv shell --python 3.8
pipenv install --dev
Build the package
pip3 install --editable .
Start the TorchServe server. You might want to do it in another window (within the same environment).
mkdir model_store
torchserve --start --model-store model_store
Once the server is running, start the module using the command:
python3 intent_detection/main.py
Note that if you run the program for the first time, the embedding model must pre-compute embeddings, and it might take several minutes. The application is ready to after the message "Application startup complete."
The Uvicorn server will run on: http://0.0.0.0:5555
To stop the server press CTRL+C and run
torchserve --stop