I have recently developed a custom rasa chat bot by following rasa tutorial. I went bit further than explained in the tutorial and created custom action which depend on a duckling server and connected a chat ui to talk to the bot. When I tried to deploy the all these together using docker-compose, I could not find a resource which directly explain how to create custom images of my chat bot and deploy it. Therefore I thought of sharing docker files which I used. I am not going through how to create a custom chat bot because that’s nicely explained here.
Once you are ready with the chat bot, we should deploy rasa core server in one docker container and action server in another.
We will first create the action server’s image by following link. We copy action file into a separate action folder and dependency will be added in the requirements file.
rasa action server docker file
# Extend the official Rasa SDK image
FROM rasa/rasa-sdk:1.6.1
# Use subdirectory as working directory
WORKDIR /app
# Copy any additional custom requirements, if necessary (uncomment next line)
COPY actions/requirements-actions.txt ./
# Change back to root user to install dependencies
USER root
# Install extra requirements for actions code, if necessary (uncomment next line)
RUN pip install -r "requirements-actions.txt"
# Copy actions folder to working directory
COPY ./actions /app/actions
# By best practices, don't run the code with root user
USER 1001
# Start the action server
CMD ["start", "--actions", "actions", "--debug"]
Now we will create the docker image of rasa server file (core + nlu). We have to update endpoints.yml with service name (the one which we add in docker-compose) of action server and we have to enable socketio in credentails.yml so that chat ui can connect to it. If you are using some other ui like Facebook or slack then we we have to enable respective lines in credentials.yml. We have to copy following files domain.yml, config.yml, credentials.yml, endpoints.yml and data etc. to the docker image and then we will train the server to create the model.
endpoints.yml
action_endpoint:
url: "http://action_server:5055/webhook"
credentials.yml
socketio:
user_message_evt: user_uttered
bot_message_evt: bot_uttered
session_persistence: true
rasa server docker file
FROM rasa/rasa:1.6.1-spacy-en
WORKDIR /app
COPY . /app
COPY ./data /app/data
USER root
RUN rasa train
VOLUME /app
VOLUME /app/data
VOLUME /app/models
USER 1001
CMD [ "rasa","shell" ]
To connect to the rasa server, I used ui provided by https://github.com/botfront/rasa-webchat. I just wrapped the script provided by them in a html file and deployed it in a nginx web server.
chat bot html file
<body>
<div id="webchat"></div>
<script src="https://cdn.jsdelivr.net/npm/rasa-webchat/lib/index.min.js"></script>
<script> // script taken from https://github.com/botfront/rasa-webchat
WebChat.default.init({
selector: "#webchat",
initPayload: "/get_started",
customData: {"language": "en"}, // arbitrary custom data. Stay minimal as this will be added to the socket
socketUrl: "http://localhost:5005",
socketPath: "/socket.io/",
title: "Title",
subtitle: "Subtitle",
params: {"storage": "session"} // can be set to "local" or "session". details in storage section.
})
</script>
</body>
chatbot ui docker file
FROM nginx
COPY index.html /usr/share/nginx/html
Finally, we will deploy everything together using a docker-compose as follows. If deploy went fine then we can access the chat ui using http://localhost/ and talk to the bot.
docker-compose.yaml
version: '3.0'
services:
rasa_server:
image: myrasaserver:v1
ports:
- 5005:5005
depends_on:
- duckling
- action_server
command:
- run
- -m
- /app/models
- --cors
- "*"
- --enable-api
- --log-file
- out.log
- -p
- '5005'
action_server:
image: myactionserver:v1
ports:
- "5055:5055"
command:
- start
- --actions
- actions
ui_nginx:
image: botui:v1
ports:
- "80:80"
duckling:
image: rasa/duckling
ports:
- "8000:8000"
References
- https://rasa.com/docs/rasa/user-guide/building-assistants/
- https://rasa.com/docs/rasa/user-guide/how-to-deploy/#building-an-action-server-image
- https://github.com/botfront/rasa-webchat