Ticket Checker

A "Ticket Checker" with Deep Learning-based intelligent support system integrated with RAG and Re-Ranking Algorithm.

Words on Config

• Set Train.train_model to true, only if you want to train the model again. I have already provided the train torch model.
• Similarly, set Train.train_tokenizer to true, only if you want to train the tokenizer again. I have already provided the train tokenizer.
• If you set the Process.Stream to True, then it will start processing tickets one by one from ticket_test: "data/data_stream_test.json" , which is 10% of the original dataset provided. This mimics 500 daily input tickets, coming to our Server. Running this takes a lot of time.
• Setting it Process.Stream to False only process songle ticket, that you define in input_query_path: "data/input_query.json".

Instructions on running the code:

Installation

1. Create a virtual environment using Python's builtin venv:

   $ python -m venv .venv
   # Linux / WSL2 / macOS
   $ source .venv/bin/activate
   # Windows
   $ .venv\Scripts\activate

2. Install poetry.

3. Install runtime and development dependencies:

   poetry install --no-root

How to run the App

You can run the application using Python or Docker. My recommended apporach is to run using Python via cmd. There are some problems with running the complete program inside a docker container.

Running with Python

To run the application using Python. Currently, it only support extrating relevant ticket for a single ticket, provided in the data/input_query.json file. Execute the following command:

python main.py

To run mlflow on the side:

mlflow ui --backend-store-uri "file:///PATH/TO/YOUR/MLFLOWRUNS/" # The path will be displayed when train is executed.

And then simply open http://127.0.0.1:5000

Running with Docker

Build the Containers

docker compose build

Start the application

docker compose up

Stop the application

docker compose down

Database Persistence

docker compose down --volumes