Deploy Rasa In Docker

Hi all, I’m new to Rasa and Docker. I have created a bot that asks user information and it stores in the database and also it fetches details from the database. I’m using postgresql for database. I want to deploy my rasa project in docker. Could anyone explain more in detail how to do that. thanks in advance. @jonathanpwheat

@Shan Are you using window, mac or Ubuntu.

@nik202 Windows

@Shan Do you have any prior docker experience?

@Shan Install docker desktop, create the rasa project in VS code. Check the rasa docker image.

@nik202 I don’t have. Anyway I try that and get back to you.

@Shan I will encourage please do some basic, beginner course or tutorial on docker, at least to know what are the process and commands and syntax used and for what process, else it will be very complex for you. This is only my suggestion, take it or leave it. Good Luck!

1 Like

Hi @nik202

Dockerfile

FROM rasa/rasa
COPY . /chatbot
WORKDIR /chatbot
RUN pip install -r requirements.txt
USER root
COPY ./actions /app/actions
USER 1001

this is my dockerfile is this correct? and I have a doubt what are libraries I should add in requirements.txt file.

@Shan No, you need to create the docker-compose file for the rasa. @Shan If you are using action.py for custom action and you using some outside libraries then you need to add that in requirenments.txt else you will see the error in logs you can add that. @Shan check this thread for your reference only Error starting action server on Docker - #5 by BELIEVEIT @Shan If further issue do let us know.

@nik202 how can I contact you privately.

@nik202 okay.do let me know when you are free

Hi @nik202. Could you explain more step by step instructions. Should I create a docker image before docker-compose?

@nik202 ?

@nik202 Don’t get me wrongly I was busy when you replied. this is my mail id shanthoshvijay16@gmail.com

what I understood for deployment from blogs and docker videos and tried like this.

First step: I have to create a docker image contains project source and requirements.

Dockerfile

FROM rasa/rasa  
COPY . /chatbot  
WORKDIR /chatbot  
RUN pip install -r requirements.txt  
USER root  
COPY ./actions /app/actions
USER 1001     

requirements.txt

pyaml  
flask   
requests  
spacy  
rasa-nlu   
rasa-core   
rasa-core-sdk  

Second step: Create Docker-compose.yml

version: "3.0"  
services:  
  rasa:  
    image: rasa/rasa:2.6.3-full  
    ports:   
      - 5005:5005  
    volumes:   
      - ./:/app  
    command:  
      - run  
      - -m  
      - models   
      - --enable-api  
      - --cors   
      - "*"  
      - --debug   
  action_server:  
    image: rasa/rasa_core_sdk:latest  
    ports:  
      - 5055:5055  
    volumes:  
      - ./actions:/app/actions  
   command:  
     - rasa  
     - run  
     - actions  

could you tell me right flow for deployment.

@Shan Is above one is running? @Shan Please delete your emailD from above post.

this is my new Dockerfile

FROM rasa/rasa
COPY . /Bot
WORKDIR /Bot
COPY ./actions /app/actions
USER 1001
CMD [“init” ]

I ran this command “docker build --tag bot .” it creates an image and I opened image in docker desktop and ran but it shows this error.

UserWarning: The model at ‘models’ could not be loaded. Error: <class ‘rasa.nlu.model.UnsupportedModelError’>: The model version is trained using Rasa Open Source 2.6.3 and is not compatible with your current installation (2.8.0). This means that you either need to retrain your model or revert back to the Rasa version that trained the model to ensure that the versions match up again.

I tried with docker-compose.yml and ran this command " docker-compose up"

version: “3.0”

services:
rasa:
container_name: rasa-server
image: rasa/rasa:latest-full
ports:
- 5005:5005
volumes:
- ./:/app
command:
- run
- --cors
- “*”
- --enable-api
- --log-file
- out.log
depends_on:
- action-server action-server:
container_name: rasa-action-server
image: rasa/rasa-sdk:latest
volumes:
- ./actions:/app/actions
ports:
- “5055:5055”

but it shows this error

rasa-server | /opt/venv/lib/python3.8/site-packages/rasa/shared/utils/io.py:97: UserWarning: The model at 'models' could not be loaded. Error: <class 'rasa.nlu.model.UnsupportedModelError'>: The model version is trained using Rasa Open Source 2.6.3 and is not compatible with your current installation (2.8.0). This means that you either need to retrain your model or revert back to the Rasa version that trained the model to ensure that the versions match up again.

rasa-server | /opt/venv/lib/python3.8/site-packages/rasa/shared/utils/io.py:97: UserWarning: Agent could not be loaded with the provided configuration. Load default agent without any model.

rasa-action-server | import psycopg2

rasa-action-server | ModuleNotFoundError: No module named 'psycopg2'

@Shan check Dockerfiles for rasa and rasa-sdk, how in that we have train the model, used requirement.txt ok you doing in right direction.

For error:

  1. you previous model is trained based on rasa 2.6.3 and your image is latest, rather train the model. with latest or change the image to 2.6.3 from latest ok?
  2. in action.py you used psycopg rather delete this import or add psycopg2 in requirement file.

I hope it make sense now.

1 Like

I updated my rasa version to latest and then I ran the command docker build --tag bot . and I ran rasa shell it works but not able to connect action server.

I’m no Docker pro, and interestingly don’t use it at all now, but I did write up a pretty super basic tutorial over a year ago about getting rasa setup with Docker. You can take a look here - Using Docker with Rasa for development - DEV Community :woman_technologist::man_technologist:

Probably doesn’t follow best practices and versions may be old, but there might be a tid-bit in there that may help figure out why you can’t talk to the action sever.

@Shan change the action-server or whatever you had mentioned in docker file in endpoint.yml from localhost → action-server. Build and try.