SurfAi Chatbot

A chatbot system that provides helpful data related to surfing conditions. The bot artificial inteligence knowledge base is built on top of Google Dialogflow and all the weather data is fetched from the Stormglass API. Google Dialogflow can be used to train machine learning classifiers to be able to detect the intent of a given question and extract important entities from it (eg: location, weather parameter).

Example interactions

How is the weather in Costa da Caparica?

Tell me how's the wave height at Carcavelos.

Forecast for Guincho.

Air temperature at Carcavelos?

REST API

Example request

curl --location --request POST 'http://127.0.0.1:8080/surf_bot/beach' \
--header 'Content-Type: application/json' \
--data-raw '{
    "text": "Swell height at Guincho",
    "sender_id": "jose"
}'

Example response

{
    "intent": "Specific Weather Parameters Intent",
    "text": "Swell height at Guincho",
    "response": "The swellHeight in Guincho will be:",
    "data": {
        "hours": [7, 8, 9, 10],
        "parameters": {
            "swellHeight": [
                1.05,
                1.06,
                1.06,
                1.08,
            ]
        }
    }
}

Supported locations

Portugal:

Carcavelos
Costa da Caparica
Guincho

Setup

This project uses Python3.7 and the Flask Framework.

Instaling all the dependencies and running the server:

$ python3 -m virtualenv venv
$ source venv/bin/activate
$ pip install -r requirements
$ python3 manage.py runserver

Required environmental variables in .env file:

SECRET_KEY=" " - random
APP_ENV=" " - development|testing|production

# Facebook
VERIFY_TOKEN=" " - token used to verify app with Facebook (random)
ACCESS_TOKEN="" - provided by Facebook

# Stormglass
STORMGLASS_URL="https://api.stormglass.io/v1/weather/point"
STORMGLASS_API_KEY=""

# Dialogflow
DIALOGFLOW_URL="https://api.dialogflow.com/v2/"
DIALOGFLOW_PROJECT_ID=""
DIALOGFLOW_CLIENT_ACCESS_TOKEN=""
GOOGLE_SERVICE_ACCOUNT_KEYS="" - path to JSON with GCP credentials

Integration with Facebook Messenger

The API supports Facebook Messenger events so that the bot can be deployed in that platform.

For this, we have to follow the steps in the Messenger Platform guide in order to create the Facebook Application and setup the webhook, pointing to /messenger_webhook.

The machine running the server has to have a valid SSL Certificate in order to be accepted by the Facebook Messenger API.

jose-correia / surf-ai-chatbot