Proof of concept implementation of Weather Chatbot using Rasa NLU and tommorow.io weather api (see References). The chat bot looks for location in user input and provides realtime weather report for given location.
Requires docker-compose, git and private weather api (https://app.tomorrow.io/home) token defined in .env file.
#contents of .env file
CLIMACELL_API_KEY=tokenRun both rasa and rasa_action_server using docker-compose
#get .env file with weather api token
docker-compose up -d --build
Now there should be running rasa API on port 5005. To test it out visit demo chatroom on localhost:8880 (cloned from https://github.com/scalableminds/chatroom)
- User asks for weather using words that are similar / close to word weather (sunny, cloudy, ...)
- Location must be specified as name of City or Country
- Spacy entity and trained DIETclassifier are weak when detecting less known cities. Needs more data or better city entity recognition.
- https://medium.com/coinmonks/rasa-python-weather-chatbot-51fc218d346d
- https://towardsdatascience.com/how-to-create-a-weather-chatbot-b8ef1b1d6703
- https://chatbotslife.com/building-a-weather-chatbot-with-rasa-nlu-and-rasa-core-8a295f29c405
- https://rasa.com/docs/rasa/
- https://docs.tomorrow.io/reference/welcome
- https://github.com/Super-Serious/bot
- https://geotext.readthedocs.io/en/latest/