AE1 Software and Technical Report for a Chat Bot

Specification

The best way to run this project is using MacOS as it has 'make' installed by default.

• Python 3.9
• Latest pip version
• Use a terminal (not Windows Cmd prompt if possible)

Usage

# Installs all dependencies (pip install)
make init

# Runs the ChatBot package
make start

# Generates documentation from docstrings (sphinx)
make build-docs

# Spins up dev server to view documentation (localhost:8000)
make host-docs

# Run unit tests
make test

# Spins up dev server to view coverage (localhost:8000)
make host-cov

Usage (detailed)

You can have a look at the Makefile to see all the commands available.

Training and running chat bot

• Install all dependencies with make init or pip install -r requirements.txt. If you are getting issues with install tensorflow on MacOS M1 chip, you'll need to follow this guide.
• If there isn't already a h5 file in chat_bot/data. Use make start-admin or python chat_bot --admin to run the project in admin mode where you can train the model with option 2.
• You can run the chat bot with make start or python chat_bot. If you experience any issue, you may have to run python chat_bot/main.py, but it is unlikely if you are using the correct version of python as per the specifications.

Unit tests

You can use make test or coverage run -m unittest discover -v chat_bot followed by coverage report -m. You can generate and host the coverage report with make host-cov or coverage html followed by cd htmlcov && python -m http.server && cd ...

If the && doesn't work on your terminal, run each command on its own line.

Open a web browser and in the url go to localhost:8000 or whatever it says in the terminal.

Documentations

You'll need to generate the documentations first. Do this with make build-docs or cd docs && sphinx-build -b html source build && cd ...

Then you can host it using make host-docs or cd docs/build && python -m http.server && cd ../...

If the && doesn't work on your terminal, run each command on its own line.

Open a web browser and in the url go to localhost:8000 or whatever it says in the terminal.

Important information

• If you evaluate the optimal number of epochs using option 3 on the admin TUI. It will change some crucial files used to run the chat bot. Please retrain the chat bot again before running the chat bot, otherwise, it will be very inaccurate.