Rasa open source 3.x Deployment method

Rasa Open Source
1

Hi dear all, I am looking for a way to deploy rasa open source only and link it to the action server via kubernetes . does anyone have a documentation or can help me ? I use a linux virtual machine

1 Like
3

@stephens anything that could help ?

4

Yes, the open source helm chart

5

Hi there. I deploy my rasa bot on k8s. here is my Dockerfile and yaml file.

Dockerfile

FROM leftsky/rasa

WORKDIR /app

COPY . .

RUN rasa train

rasa.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: rasabot-deployment
spec:
  replicas: 1
  selector:
    matchLabels:
      app: rasabot-deployment
  template:
    metadata:
      labels:
        app: rasabot-deployment
    spec:
      containers:
      - name: rasabot-deployment
        image: "the docker image you have built"
        env:
        - name: APP_NAME
          value: RASABOT
        - name: APP_VERSION
          value: 0.0.1

        ports:
        - containerPort: 9100

After this you can exec kubectl apply -f rasa.yaml to deploy your rasa bot.

By the way.The image used above leftsky/rasa is build by myself.You can build yourself one.Here is its needed files.

Dockerfile

FROM python:3.7.9-slim

RUN apt update
RUN apt install supervisor -y

RUN python -m pip install --upgrade pip
RUN pip install rasa
RUN pip install jieba
RUN pip install paramiko

COPY start.sh /start.sh
RUN chmod +x /start.sh
COPY supervisord.conf /etc/supervisord.conf

EXPOSE 9100
WORKDIR /app

CMD ["/start.sh"]

start.sh

#!/bin/bash

# ----------------------------------------------------------------------
# Start supervisord
# ----------------------------------------------------------------------
exec /usr/bin/supervisord -n -c /etc/supervisord.conf

supervisord.conf

[unix_http_server]
file=/dev/shm/supervisor.sock

[supervisord]
logfile=/tmp/supervisord.log
logfile_maxbytes=50MB
logfile_backups=10
loglevel=warn
pidfile=/tmp/supervisord.pid
nodaemon=false
minfds=1024
minprocs=200
user=root

[rpcinterface:supervisor]
supervisor.rpcinterface_factory = supervisor.rpcinterface:make_main_rpcinterface

[supervisorctl]
serverurl=unix:///dev/shm/supervisor.sock

[program:rasa-actions]
command=rasa run actions
autostart=true
autorestart=true
priority=10
stdout_logfile=/dev/stdout
stdout_logfile_maxbytes=2048
stderr_logfile=/dev/stderr
stderr_logfile_maxbytes=2048

[program:rasa-api]
command=rasa run --enable-api --port 9100 --cors *
autostart=true
autorestart=true
priority=10
stdout_logfile=/dev/stdout
stdout_logfile_maxbytes=2048
stderr_logfile=/dev/stderr
stderr_logfile_maxbytes=2048

Hope this helps.

6

Hello thanks, do you have a video ressource ?

7

hello welcome sir

i will take a look and i will tell you thanks again

it assumes that you have a dockerhub right ?

8

Yes, leftsky/rasa is on docker hub

9

ok for me i dont have a docker hub i think i will create one locally

for the rasa models if mine has been built locally how can i link it with the helm ?

10

hi @stephens , seems like my github integration (or dwnload file ) in order to train an initial model with the deployment of rasa open source via helm is not working i have tried all the recommandations but the files are not beeing download how can i do ?