fossbot-text-to-cmd

Starter project for the FOSSBot4AI course - a small Python program that classifies short English movement commands ("go forward", "turn left", "stop", ...) into discrete actions and maps each action to wheel speeds for the FOSSBot.

This repository is the starting point for two labs:

• Lab 3 - Introduction to Python: external libraries (including AI). Fill in the TODOs in src/classifier_sklearn.py (TF-IDF + LogisticRegression) and src/classifier_st.py (sentence-transformer cosine similarity). src/wheel_mapping.py is provided pre-filled - it is the lookup table from action label to wheel speeds.
• Lab 4 - Dockerization + VSCode. Containerise the working application by writing your own Dockerfile and docker-compose.yml, then run it from the command line and from a VSCode dev container.

Layout

src/                          Python sources (classifiers + entry point text_to_wheels)
data/training_commands.csv    Labelled dataset for the sklearn classifier
data/examples/                Sample command files to classify
_solutions/                   Reference solutions: Lab 3 classifiers + Lab 4 Dockerfile, compose, devcontainer
requirements.txt              Python dependencies

Quick run (after the Lab 3 TODOs are completed)

pip install -r requirements.txt
python -m src.text_to_wheels \
    --input data/examples/basic.txt \
    --output basic_sklearn.json \
    --classifier sklearn

Replace --classifier sklearn with --classifier st to use the sentence-transformer variant.

Licence

CC BY 4.0.