Run in docker

Make sure your host has docker service installed.

1. Make sure you've built the rasa3-for-botfront docker image already.
2. Build the botfront docker image.

git clone https://github.com/djypanda/botfront-for-rasa3.git
cd botfront-for-rasa3
docker build . -t bf-for-rasa3:v0.1

If your building process exit on OOM error, just increase the docker service resource memory(e.g. 8G), and try again.

This container can use 5944M memory in total.
#16 0.350     If it aborts with an out-of-memory (OOM) or ‘non-zero exit code 137’ error message,

3. Create the project root directory.

If you haven't setup the dev environment, you should install botfront module: npm install -g botfront

Create the project somewhere not in the source code tree.

botfront init

4. Modity configuration in the project directory.

• edit .botfront/docker-compose-template.yml

Original	Change to
version: "3.0"	version: "3.9"
botfront:   image: botfront/botfront	botfront:   image: bf-for-rasa3:v0.1
rasa:   build:     context: ./rasa   args:     RASA_IMAGE: ${IMAGES_CURRENT_RASA}   ...   volumes:     - ./models/:/app/models	rasa:   image: rasa3-for-bf:v0.1   ...   volumes:     - ./models/:/app/models     - ./config/:/app/config

• edit .botfront/botfront.yml

images:
  default:
    botfront: 'bf-for-rasa3:v0.1'
    rasa: 'rasa3-for-bf:v0.1'
    duckling: 'botfront/duckling:latest'
    mongo: 'mongo:6.0.1'
    actions: 'rasa/rasa-sdk:3.2.1'
  current:
    botfront: 'bf-for-rasa3:v0.1'
    rasa: 'rasa3-for-bf:v0.1'
    duckling: 'botfront/duckling:latest'
    mongo: 'mongo:6.0.1'
    actions: 'rasa/rasa-sdk:3.2.1'

5. Start services.

botfront up

NOTE - After the first run you can use docker-compose up [-d] to start those services either.

6. Setup chitchat project

• Create a chat project, follow the link below:

  • https://botfront.io/docs/installation/local-machine

• rasa need a trained nlu module to startup (They don't have one in the init project root/models folder)

  • !!! Copy a dummy one from demo_project/models

• This step is critical, botfront set-projects chitchat-5v7YtAmFF (Change to your project name).

  • https://botfront.io/docs/installation/local-machine#connect-rasa-to-your-new-project

7. Open http://localhost:8888/ in browser.

First time it will take a quite long time to download meteor.js

8. Setup the zh language

In 'nlu' menu -> 'Settings' tab:

Update the pipeline with below text:

pipeline:
  - name: SpacyNLP
    model: zh_core_web_md
  - name: SpacyTokenizer
  - name: SpacyEntityExtractor
  - name: SpacyFeaturizer
  - name: RegexFeaturizer
  - name: LexicalSyntacticFeaturizer
  - name: CountVectorsFeaturizer
  - name: CountVectorsFeaturizer
    analyzer: char_wb
    min_ngram: 1
    max_ngram: 4
  - name: DIETClassifier
    constrain_similarities: true
    epochs: 100
  - name: EntitySynonymMapper
  - name: ResponseSelector
    epochs: 100

9. Train exampes, after train finished, click the chat icon(at right bottom), talk to your chatbot!

10. That's all, have fun!

Development Installation

1. Botfront is a Meteor app, so the first step is to install Meteor
2. Then clone this repo and install the dependencies:

git clone https://github.com/djypanda/botfront-for-rasa3.git
cd botfront-for-rasa3/botfront
meteor npm install

3. Install the CLI from the source code:

# if you installed Botfront from npm uninstall it.
npm uninstall -g botfront
# Install the cli from the source code
cd cli && npm link

Botfront needs to be connected to other services, especially Rasa.

1. Create a Botfront project with botfront init (somewhere else, not in the repo) Make sure your host has docker service installed. 'botfront init' command will create a instance project directory and download pre-build docker images in which comes with rasa2.

2. Start your project with botfront up -e botfront -e rasa. This will run all services except the Botfront app and rasa since you are going to run it with Meteor and rasa 3 locally and mannually.

3. Make sure you started rasa3 service (rasa3-for-botfront, the modified version).

4. Go back to the botfront checkout cd botfront-for-rasa3/botfront and run Botfront with meteor npm run start:docker-compose.dev. Botfront will be available at http://localhost:3000 so open your browser and happy editing 😸

What's new ???

The ultimate target of this copy is to decouple the botfront and rasa, let botfront only do the editing work, and keep rasa as untoucned as possible!

[DID:]

1. Removed multi-language support in one rasa instance. This this also suggested officially to be achieved multi-instances.
2. Add chinese as default language.
3. Temporary removed Gazette feature and some other features that need rasa to change the code inside nlu module.
4. Commented Story Behaviour

[TODO:]

1. Action can be edited in botfront
2. Change the train payload from json (deprecated!) to yaml format
3. Make the botfront cli work for this repo
4. TBA.

Original instruction goes here !!!