chatwoot/rasa-agent-bot-demo

Archiving the Repo

Moving this to Chatwoot Implementation Samples.

Chatwoot Agent Bot Demo using Rasa

You should be running this demo on a local installation of Chatwoot and a non dockerised setup for the localhost and ports to be accessible for all the services involved. If you are intending to run in a remote server, ensure you change the localhost urls with appropriate IP addresses and make sure the ports should be accessible for all the services involded.

This is a sample implementation of agent bot capabilities in chatwoot using rasa . Rasa Open Source is a machine learning framework to automate text- and voice-based assistants.

You can refer the rasa documentation to get it up and running in your machine.

This implementation isn't a recommended set up for production, but just to illustrate the capabilities of the platform. Please build on top of this ideas discussed to have in running in production.

Follow the given steps to get your agent bot integration up and running.

Refer the Video Walkthrough and blog post

Get a rasa project up and running.

Go to a new directory and create a rasa project. If you have rasa installed in your machine you can get it up and running by follow in commands. Refer docs to get the installation up and running.

mkdir rasa
cd rasa
rasa init --no-prompt

go to credentials.yml file in the directory and ensure the following value is set. This is to ensure we can communicate with rasa through rest api

rest:
  # you don't need to provide anything here - this channel doesn't
  # require any credentials

start the rasa server with following command

 rasa run -m models --enable-api --log-file out.log

Get your chatwoot up and create an agent bot

go to your chatwoot directory and ensure your local server is running. Start a rails console in your directory.

bundle exec rails c

Inside the rails console, type the following commands to create an agent bot and get its access token. Save the retrieved token as you would need it in further step.

bot = AgentBot.create!(name: "Rasa Bot", outgoing_url: "http://localhost:8000")
bot.access_token.token

Connect Agent Bot to your inbox by running the following command

AgentBotInbox.create!(inbox: Inbox.first, agent_bot: bot)

Clone this repo into your machine and run the rasa router script.

clone repo using the following command.

git clone [email protected]:chatwoot/rasa-agent-bot-demo.git

Using Python

open up the python file in your editor and change the follow values with appropriate ones.

rasa_url, chatwoot_url and chatwoot_bot_token.

Then run pip install -r requirements.txt and python3 -m gunicorn --workers=1 test:app -b 0.0.0.0

Using PHP:

open up the rasa-router/index.php file in your editor and change the follow values with appropriate ones.

$rasa_url = 'http://localhost:5005';
$chatwoot_url = 'http://localhost:3000';
$chatwoot_bot_token = '<your agent bot token>';

run the php server in the rasa-router directory

cd rasa-router
php -S localhost:8000

Connect to your chatwoot webwidget and start a conversation.

if you are on your local machine, you can access the widget through the test page

http://localhost:3000/widget_tests

Notes

You can also refer to the RasaHQ / rasa-demo for adding additional capabilities to your bot. If training rasa through scripts isn’t your thing, check the exciting rasa projects which gives a UI to create your rasa stories.

• Botfront
• Articulate

You can build on top of the ideas discussed here to implement your solutions. Refer to the chatwoot api to see the available options in chatwoot for your bots.Pretty excited to see what you guys come up with.