Install Clai 2.0 on Microsoft Windows

The Windows installation is only included with the Enterprise Edition.

This document is for installing Clai 1.2.x and assumes the packages will be installed into directory \app

Locate the installation files in \app

Installation of Rasa and Action Services

Install Python, the Microsoft Visual C runtime and Redir using their respective installers

  • Install python 3.7.9
  • Install VC_redist.x64.exe
  • Install Redis-x64-3.2.100.msi

Note: whilst installing redir, do put it in the app directory and most likely you’ll not need to add a firewall exception unless running in a cluster in which case the exception should be edited to limit who can connect to redis and also enabling a password for access to the redis service.

This document assumes that you have tar available in your Windows environment. If not you should use a tool such as 7-Zip or alternate in its place.

In a command window run as Administrator:

cd \app

tar xf actions-1.0.3.tar.gz

mkdir models

mkdir logs
  
python -m venv rasa
	
cd rasa\lib
	
tar xf ..\..\site-packages-1.0.3.tar.gz

If the intention is to use Spacy for NLU, then also install the relevant language pack, i.e. english.

cd ..\..
	
rasa\scripts\python -m spacy link en_core_web_md en

Now disable Rasa telemetry

rasa\scripts\python -m rasa telemetry disable

Install the services:

For Clai Server and Actions services:

cd \app
set APP=c:\app

Check the script contents are correct and then run the scripts

svc-server.cmd
svc-actions.cmd

Start the services

Installation of Clai Admin and Mongo services

Install NodeJS and MongoDB using their respective installers

  • Install node-v12.21.0-x64.msi
  • Install mongodb-windows-x86_64-5.0.4-signed.msi

In a command window run as Administrator:

cd \app

mkdir models

tar xf admin-1.2.x.tar.gz

Install the Clai admin service:

For the Clai Admin Service:

cd \app

set APP=c:\app

set ADMIN_ROOT_URL=http://HOST:8888

Check the script content, correct it and then run the script

svc-admin.cmd

Installation of Nginx proxy and ssl offload

Unzip the file nginx-1..0.zip

cd \app

unzip nginx-1.18.0.zip

Rename the directory to just nginx

ren nginx-1.18.0 nginx

Edit as required then copy the nginx.conf file to nginx\conf\nginx.conf

copy nginx.conf nginx\conf\nginx.conf

For nginx supporting rasa then also copy the file Clients\index.js to nginx\html

copy Clients\index.html nginx\html\index.html

copy Clients\index.js nginx\html\index.js

Completion

Add the necessary rules to the windows firewall for the claisvr and claiadmin ports.

Clean up

You can now delete the following files used as part of the installation or move them to a backup directory:

  • python-3.7.9-amd64.exe
  • VC_redist.x64.exe
  • Redis-x64-3.2.100.msi
  • node-v12.21.0-x64.msi
  • mongodb-windows-x86_64-4.4.4-signed.msi
  • nginx-1.18.0.zip
  • admin-1.1.2.tar.gz
  • site-packages-2.3.3.cl.5.tar.gz
  • svc-server.cmd
  • svc-actions.cmd
  • svc-admin.cmd
  • svc-web.cmd
  • nginx.conf

Necessary changes to be considered in Rasa and Admin

To setup services to the clai you have to define various configuaration settings as shown below:

Rasa server

For Rasa server, add the below variables:

  • CLAI_ENV : Add service name
  • ADMIN_URL : Admin graphql URL

Admin server

For admin server, add the below paths:

  • CLAIENVSPATH : path for the file in which services are defined
  • CLAIMODELSPATH : path where models will be stored and retrieved after training
  • ADMIN_DB_URL : add the mongo db url including the database name

Setting up services

  • For creating services you have to create clai_services folder and create a file called claiservice.yaml and specify credentials, endpoints and instance. Path of the file would be provided in enviornment variable named “CLAIENVSPATH”

  • Once the rasa services are created, then you are able to associate those services to a particular project in clai according to the requirement.

  • You can create multiple rasa services with unique names and assign them to projects in Clai admin settings.

  • Service names can be named according to your preferences and it can be configured from one project to other and vice versa.

  • As Clarity uses CORS you will need to enable CORS within the webchat channel to enable Clarity to connect to the channel as doing so you will enable continued use of the standalone and admin based webchat clients.

  • Service name should be passed in the endpoints while configuring model server. Endpoints let you define how your Rasa instance communicates with Clai and the actions server.

    • the Clai API to query the bot responses (nlg)
    • the actions server (action_endpoint)
    • the tracker store (tracker_store)

See the below example for the services.

SERVICE : ## name of the service
  credentials:
    rasa_addons.core.channels.webchat.WebchatInput:
      session_persistence: true
      base_url: http://CLAIHOST:CLAIPORT
      socketio_path: /development/socket.io/
      socket_path: /development/socket.io/
      cors_allowed_origins:
        - http://CLARITYHOST:CLARITYPORT
        - http://CLAIHOST:CLAIPORT
  endpoints:
    models:
      url: http://CLAI_HOST:CLAI_PORT/models/SERVICE
      wait_time_between_pulls: 100 # [optional](default: 100)
    nlg:
      type: rasa_addons.core.nlg.GraphQLNaturalLanguageGenerator
      url: http://CLAI_HOST:CLAI_PORT/graphql
    action_endpoint:
      url: http://ACTION_SERVER:ACTION_SERVER_PORT/webhook
    tracker_store:
      store_type: rasa_addons.core.tracker_stores.botfront.BotfrontTrackerStore
      url: http://CLAI_HOST:CLAI_PORT/graphql
  instances:
    host: http://RASA_HOST:RASA_PORT

  • All services would be listed in the service page and you can see the available services and their associated project with enviornment details.

  • You can also add a service with different environments in a projects page such as development, training and production service.

  • One service can be re-used in many projects for training environment but can’t give same service name for both development and production environments.

You must now restart the rasa service, i.e. rasa-development. Now test by accessing:

  • Clarity at http://CLARITYHOST:CLARITYPORT/pmb
  • the standalone webchat at http://CLAIHOST:CLAIPORT/webchat
  • the Admin webchat within the project in the admin tool at http://CLAIHOST:CLAIPORT

If there is no path added by the client in the “admin > settings > webhooks > upload model” then the model will be stored locally in the path provided inside enviornment variable “CLAIMODELSPATH”

All the models stored in the model server path would be visible in Models page inside admin. You can view/delete models and review their associated projects.

Uninstall Clai for Windows

Uninstall of Rasa and Action Services

In a command window run as Administrator stop the services and then remove them:

net stop claisvr

net stop claiactions

net stop claiweb

cd \app

nssm remove claisvr

nssm remove claiactions

nssm remove claiweb

Remove the installation directory

cd \app

rd /S /Q rasa

rd /S /Q actions

rd /S /Q nginx

Uninstall python 3.7.9 in Apps & Features control.

Uninstall of Admin and Mongo services

In a command window run as Administrator stop the services and then remove them:

net stop claiadmin

net stop claiweb

cd \app

nssm remove claiadmin

nssm remove claiweb

Remove the installation directory

cd \app

rd /S /Q admin

rd /S /Q nginx

Uninstall NodeJS, MongoDB and Redis in Apps & Features control.

Completion

If all components have been removed from the server now delete \app to remove the model and logs directories.

cd \

rd /S /Q \app

"Install Clai 2.0 on Microsoft Windows" was last updated on: 1 January, 2023