envirocar/envirocar-rasa-bot

enviroCar Rasa Bot

A bot for enviroCar android application part of Google Summer of Code 2022 project: voice command.

👇 Prerequisites

Before installation, please make sure you have already installed the following tool:

• pipenv

pip install pipenv

We will use pipenv to easily manage and setup a working environment.

🛠️ Installation Steps

Once you have cloned the project follow these steps to install:

• Initialize a virtual environment and install dependencies via use pipenv

cd envirocar-rasa-bot
pipenv install

🧑‍💻 Training and Testing

• Train bot

rasa train

• Test bot on terminal

rasa shell

• start rasa server and test locally

rasa run --enable-api --port 5005

• start rasa server to use the custom channels

rasa run --credentials credentials.yml --enable-api --port 5005

• start actions server

rasa run actions -p 5055

Once the server is up and running, test the bot directly via postman
Here's a blog to explore the rasa apis with postman

But here are some of the common apis to use:

1. Testing the bot
   Send a POST request to a particular channel http://localhost:5005/webhooks/<channel_name>/webhook with the body. This project provides 2 channels.
   
   a. rest channel
   
   This Channel is given by the rasa itself, and we cannot add extra functionalities to it. E.g. We cannot send extra data like metadata in the request.
   
   b. envirocar channel
   
   This Channel is a custom channel created to use the extra functionalities and send extra data like metadata or some credentials in the request.
   
   Learn more about the Custom Channels from here

{
    "message": <your message>
}

2. Check rasa version A GET request to http://localhost:5005/version without the body.

You could also explore the apis with postman via video.

envirocar-rasa-bot's Issues

Enhance documentation

Description

The documentation within the README.md explains to set up everything and start the Rasa servers. However, I miss some example requests, for triggering the stories. Especially, the extra metadata, which is required for calling the Action server is hard to guess. It would be great if the documentation provided some examples for it and explain the model request and response model schema briefly.

This issue was transferred from older repository to here.

cc: @SebaDro

Setting up CI/CD for testing the assistant

Description

Even though developing a contextual assistant is different from developing traditional software, we should still follow software development best practices. Setting up a Continuous Integration (CI) and Continuous Deployment (CD) pipeline will ensure that every commit/PR is not introducing new problems or negatively impacting the performance of our assistant.

Solution

We can use GitHub Actions to run CI checks on every PR or commit. This action will train, test, and publish a summary.