How howl-anderson/weatherbot works
Project type and purpose
• This is a Rasa‐based chatbot (“WeatherBot”) that answers Chinese weather queries.
• It’s architected as four isolated “micro‐apps” (UI, Dialog Manager, NLU, Action Server) that collaborate at runtime.
File structure
• The code for each component lives in its own repository, indicating a service‐oriented / microservices–like architecture rather than a monolith.
Extract key architecture details from the README
• Components:
• User Interface (WeatherBot_UI) – web frontend or any chat channel via Rasa Channels
• Dialog Manager (WeatherBot_Core) – Rasa Core for slot‐filling, story handling
• NLU (WeatherBot_NLU) – Rasa NLU for intent classification & entity extraction
• Action Server (WeatherBot_Action) – custom actions that call the external weather API
• External dependency: “心知天气” REST API
• Communication style: HTTP/REST (Rasa’s built‐in HTTP endpoints / webhooks)
Components & their interactions
a. User → UI
• UI sends user messages to Rasa Core via a Channel adapter (HTTP webhook).
b. Rasa Core (Dialog Manager)
• Receives message → calls Rasa NLU (HTTP) → gets intent/entities → runs dialogue policy →
• If a custom action is needed → invokes Action Server (HTTP) → Action Server calls weather API → returns weather data → Core constructs bot reply
c. Rasa Core → UI
• Core returns the reply to the UI via the same channel endpoint
d. External Service
• Action Server ↔ Weather API
Architectural patterns & principles
• Component isolation (microservices–style)
• Clear separation of concerns: NLU vs dialogue management vs business logic
• HTTP/REST interfaces between all components
• Slot–based frame dialogue management
Technologies / frameworks
• Rasa Core (Python)
• Rasa NLU (Python)
• Rasa Core SDK (Action Server, Python)
• Web frontend (any framework; in this repo it’s WeatherBot_UI, likely Node/React or Flask/Jinja)
• External Weather REST API (心知天气)
• Docker containers (recommended)