carlobee/RoboCafe

RoboCafe

A chatbot built in RASA to sit beside a coffee machine and assist users in making their beverages!

Running the bot

You need to run Furhat SDK, ngrok, and Rasa in order to run the bot.

First run ngrok on http 5005:

ngrok http 5005

Then change the following lines in Furhat SDK interaction.kt file with your ngrok url (in the picture):

val RASA_URL = "\'http://<ngrok_url>/webhooks/myio/webhook\'"

Then build the Furhat skill.

After, run Rasa using this command from within the RoboCafe directory on different terminal windows:

rasa run

Note that every time you run ngrok you will get a new url, so you will have to build and update the Furhat Skill

Building Furhat skills

In order to run the bot you need to have the Furhat SDK.
The Furhat SDK uses skills to run that need to built in order for them to run.
Check out the docs and this video for more details on how to build a skill.

Rasa and Rasa X Installation

In order to run the Rasa bot we encourage you to use the Alana Anaconda environment found in this repo.
You can do this by running the alana_installation.sh in the Anaconda prompt (on Windows) or terminal (MacOS / Unix) or by using the terminal in pycharm. Once created activate it with conda activate Alana, you will see (Alana) at the left of your prompt.

Rasa

For testing we recommend installing Rasa

pip install rasa

Rasa X

For development we recommend installing Rasa X.
This gives more features for training the bot.

pip install rasa-x --extra-index-url https://pypi.rasa.com/simple

If this installation is very slow try downgrading your pip installation to 20.2 with: pip install --upgrade pip==20.2

You may have to install Visual C++ build tools which consist of 3GB+ of files.
This is not required in normal Rasa but is for Rasa X

Cloning repo

This can be done with the GitHub desktop app or with the following command in the terminal (MacOS / Unix).

git clone https://github.com/carlobee/RoboCafe

Running Rasa

Before running rasa make sure to run the actions server in a separate prompt or terminal window/tab.

rasa run actions

You can then run either of the following:
rasa shell runs a command line chat interface.
rasa interative runs a command line debug interface.

Running Rasa X

Same as Rasa but you can run an interactive web interface by running:

rasa x

This allows you to train your bot in real time and save different models.

Telegram server setup

This requires ngrok, which creates a unique URL that can be used to access your bot server outside your local network.
Once you have downloaded ngrok navigate to the folder in your prompt/terminal and run the command ngrok http 5005

The window should display a forwarding URL (e.g. https://6e28a3215a9d.ngrok.io -> http://localhost:5005),
copy the first address (https://6e28a3215a9d.ngrok.io) to the webhook_url field in the credientials.yml file.

webhook_url: "https://<your-ngrok-url>/webhooks/telegram/webhook"

Open a new terminal/prompt window or tab and use the command rasa run to run the bot

Now open your Telegram app and add the user @robot_cafe_bot to start chatting!

Development

When pushing changes don't push to main branch. Create your own branch or push to the branch for your team (e.g. RASA).

Pull requests will be reviewed by the whole team before merging with the master branch