Part 1 - getting started with your first Rasa bot

Introduction

Rasa is an open-source chatbot framework to create and automate conversational applications. It provides several features to make you efficiently develop your bot. With Rasa, you can generate your training data, train your model, run tests, and connect your bot to APIs and different messaging channels.

Prerequisites

Don't take the following prerequisites too seriously, but it is preferable to know how to use:

• Terminal, to move around your computer's hard drive.

• Anaconda is required.

• Git fundamentals (I'm kidding, you need to run only one command to retrieve Rasa open source code).

Building from source

Because you are a software engineer, it is preferable to use the development version of Rasa open source. You can follow the instructions below to set up Rasa properly in your local machine.

Create a Conda environment

First, let's create a new Conda environment based on Python 3.7 because Rasa and Tensorflow are more compatible with this particular version according to Rasa Documentation. Then, we can use the poetry package manager to install the dependencies needed to run Rasa properly.

$ conda create -n rasaenv 'python>=3.7,<3.8' poetry

Note that you must activate your new Conda environment to install Rasa and its dependencies in the right place. To activate a Conda environment, use the following command:

$ conda activate rasaenv

Clone the Rasa source code

Create a new directory in your local hard drive where you can clone the Rasa source code. From your terminal, don't forget to go to the directory that was just created:

(rasaenv):~$ mkdir ~/code
(rasaenv):~$ cd ~/code
(rasaenv):~$ git clone git@github.com:rasahq/rasa
(rasaenv):~$ cd ~/code/rasa 

Run the poetry package manager

To install rasa in the right environment, make sure your Conda environment is activated to run the poetry package manager. Poetry comes with a bunch of python packages like Tensorflow that Rasa needs for training.

(rasaenv):~$ poetry install

Note that some machine learning algorithms require additional python packages to be installed, this page on Rasa documentation can help you configure those dependencies.

Create a Rasa project

Any Rasa project contains some YAML files that form the basis of the skeleton conversation design of any bot. So, you can create a new directory in where you cloned the Rasa source code to hold your Rasa chatbot projects:

(rasaenv):~$ mkdir ~/code/rasa-projects
(rasaenv):~$ cd ~/code/rasa-projects

To initialize your Rasa project, the rasa init command comes with a list of questions to enter a path for your project and train your initial model. After you run the mentioned command, give a name to your project at the first prompt so Rasa can create a structure for your chatbot. Then, you can accept all the prompts except the latest one in case you don't want to run your bot at the moment.

? Please enter a path where the project will be created [default: current directory] my-first-rasa-chatbot                   
? Path '/home/code/rasa-projects/your-first-rasa-chatbot' does not exist 🧐. Create path?  Yes 

Created project directory at '/home/code/rasa-projects/my-first-rasa-chatbot'.
Finished creating project structure.
? Do you want to train an initial model? 💪🏽  Yes                                                                             
Training an initial model...
The configuration for policies and pipeline was chosen automatically. It was written into the config file at '/home/code/rasa-projects/my-first-rasa-chatbot/config.yml'.
Training NLU model...
...
Core model training completed.
Your Rasa model is trained and saved at '~/code/rasa-projects/my-first-rasa-chatbot/models/20220121-114355-burgundy-moat.tar.gz'.

? Do you want to speak to the trained assistant on the command line? 🤖  (Y/n) No

Ok 👍🏼. If you want to speak to the assistant, run 'rasa shell' at any time inside the project directory.

Start with Rasa playground data

When you run the run init command, Rasa created a project structure for your bot that includes some dependency files like config.yml and some directories including data and models. So, let's create a new YAML inside data directory, and name it something like my-first-training-data.yml.

(rasaenv):~$ cd ~/code/rasa-projects/my-first-rasa-bot/data
(rasaenv):~$ touch my-first-training-data.yml

Rasa has a nice conversation design that can get you to start creating the first training data for your bot. You can go to Rasa Documentation Playground to build your assistant bot with an interactive guide. To keep things simple at the moment, let's copy the content of this file immediately into your training YAML file. Then, don't forget to import your intents in the domain.yml as we did here.

Train your bot

Next, to run your first model using the training data you just created, you can use the rasa train command that comes with two attributes. The --data is used to enter the path where the training data is located, and the --config to define the config.yml that your model will use to make predictions based on user inputs.

(rasaenv):~$ cd .. 
(rasaenv):~$ rasa train --data data/my-first-training-data.yml  
--config config.yml

The training command above will help your Rasa bot classify what the user is saying into the right intents, in a way your bot will be able to recognize different user inputs even with words that are not included in your training data file. The training process will be ended up by saving your model to the models directory of your project.

Run your bot

Lastly, you can run your first Rasa bot using the rasa shell command. Try to test manually your bot by typing some inputs that are not the same as you typed into the training YAML file. Feel free to stop the conversation at anytime you want using the /stop command.

What’s next?

I hope this tutorial could help you build your first basic Rasa bot. In the next parts, we will take a closer at the structure that Rasa created for our first chatbot to understand each component and what is used for. Later in this series, we will confine our bot to a specific topic and write some tests to test the accuracy of our bot.

If you have any questions, leave them in the comments below, or feel free to contact me on LinkedIn. Don’t forget, this post is public, so feel free to share it.