TorBot Files

This release makes the page classifier much cheaper to use during crawls.

What's changed

• Trains the NLP classifier lazily from the packaged CSV and caches it for the process.
• Stops retraining the model for every crawled page.
• Returns prediction confidence for the selected category instead of the old misleading random-split accuracy value.
• Keeps normal classification from generating a training_data/ directory at runtime.
• Removes process-wide directory changes from the NLP helpers.
• Adds direct NLP coverage for text extraction, CSV loading, classifier scoring, and path-safe training-data export.

Install

:::sh
pip install torbot==4.3.0

Checks

• .venv/bin/python -m pytest tests/test_nlp.py
• .venv/bin/python -m pytest
• .venv/bin/python -m flake8 src/torbot/modules/nlp/main.py src/torbot/modules/nlp/gather_data.py tests/test_nlp.py
• scripts/publish.sh