How hellohaptik/chatbot_ner works
Project type and purpose
• This is an open-source, Python-based Named-Entity-Recognition (NER) API framework tailored for conversational AI (chatbots).
• It exposes HTTP endpoints (via Django/ASGI) to accept text inputs, runs a pipeline of detectors (numeral, pattern, temporal, textual), and returns extracted entities.
• It also provides CLI commands to manage/support its Datastore (Google Cloud Datastore or Elasticsearch) and ships with Docker setups for development and production.
File structure
• Top-level “chatbot_ner/” directory is a Django project (asgi.py, wsgi.py, settings.py, urls.py, manage.py).
• “ner_v1/” and “ner_v2/” contain detector implementations (v2 for numeral, temporal, pattern with multi-language support; v1 for textual lookups).
• “datastore/” holds code to create/populate/query Google Datastore or Elasticsearch.
• “external_api/” wraps any third-party API calls (e.g., dictionary lookups).
• “lib/nlp/” contains common NLP utilities (tokenizers, lemmatizers, CRF models under data/models).
• “dev_docker/” and “docker/” define separate Dockerfiles and compose setups (development vs production with Nginx + Supervisor).
• Configuration and logging/Sentry setup under “config.py”, “setup_logs.py”, “setup_sentry.py”.
• “docs/” contains architecture sketches (entity hierarchy), install/API docs, and images.
README for additional insights
• Confirms four detector categories (numeral, pattern, temporal, textual) and their language support.
• Shows high-level entity hierarchy.
• Mentions API endpoints (detailed in docs/api_call.md).
• Points out separate deployment configs for Docker and contribution patterns.