LLMs-by-me

A collection of chatbots built while learning, from exact phrase lookup to transformer-based text generation.

The progression

File	Approach	Notes
N1-mini.py	Exact phrase lookup	JSON-backed chatbot with a Tkinter GUI
N1-violet.py	TF-IDF + cosine similarity	Adds approximate matching and aliases
N2.py	TF-IDF + cosine similarity	Expanded alias matching and debug output
N3-mini.py	TF-IDF + cosine similarity	Later chatbot variant; loads mass_training.json when present
N3-o.py	TF-IDF + cosine similarity	Expanded aliases; loads mass_training.json when present
N4-max.py	DialoGPT text generation and fine-tuning	Uses Hugging Face Transformers; choose training data in the GUI
chatbot_data.json	Data	Chat and training examples for the JSON-backed versions

Requirements

• Python 3
• Tkinter for the desktop interfaces (included with many Python installations)
• scikit-learn for N1-violet.py, N2.py, N3-mini.py, and N3-o.py
• torch and transformers for N4-max.py

Install the machine-learning packages you need for the script you plan to run. PyTorch installation options vary by operating system and hardware; see the PyTorch install guide.

Usage

Run any script from the repository directory. For example:

python N2.py

N4 uses microsoft/DialoGPT-medium by default. You can pass another Hugging Face model ID or a local model directory:

python N4-max.py --model microsoft/DialoGPT-medium

Click Train Model to select a JSON file containing Q&A objects with input and response fields. N4 saves the fine-tuned model and tokenizer to ./fine_tuned_model; run it later with:

python N4-max.py --model ./fine_tuned_model