How alexander1022/darwin works
Project type and purpose
• Darwin is a Python-based NLP/ML toolkit and conversational dialog trainer for Bulgarian.
• It’s structured as a library/CLI tool rather than a full-stack web app. It ingests dialogues, uses SpaCy (and eventually simplemma) to parse language, manages conversation state, and supports training and evaluation.
File structure
• Top-level directories:
• additional/ ← JSON resources (intents, characteristics)
• darwin/ ← core Python package modules
• tests/ ← unit tests for comparison & search logic
• Key modules in darwin/:
• darwin.py ← entry point / orchestrator
• chatbot.py ← chat I/O interface
• training.py ← model-training logic
• context_parser.py ← dialogue context extraction
• memory_manager.py ← conversation state storage
• contact_manager.py ← user/contact data interface
• search.py ← text-search utility
• comparisons.py ← intent/entity comparison logic
• tags.py ← linguistic tagging helpers
• tools.py ← shared utilities
• Configs & data:
• .env.template ← environment variables stub
• contacts.csv ← sample contact data
• additional/*.json ← intent definitions, Bulgarian characteristics
• Dependencies: defined in requirements.txt (e.g., spaCy).
Key insights from README
• Uses SpaCy with a Bulgarian model; future simplemma integration.
• Devised as a Python package installable from source.
• Core functionality: dialogue handling, training conversational models.