Soft ware specifications and other questions!

Rasa Open Source
1

Inquiry about Rasa Open Source Project

Hi,

I hope this email finds you well. I am currently utilizing the Rasa Open Source Python library to develop a chatbot for a project focused on the Korean language. I have a few questions regarding Rasa that I hope you can assist me with.

To provide context, my project involves a server built using the HTML Flask framework in Python, with the client-side implemented using HTML and JavaScript. The Rasa library versions I’m working with are as follows:

rasa: 3.6.2 rasa-core: 0.8.6 rasa-nlu: 0.11.5 rasa-sdk: 3.6.1 In my project, the last rule and action are only set to ‘action_default_fallback’, and I aim to distinguish between intents in the ‘credentials.yml’. Below is the ‘config.yml’ file I am currently using:

yaml Copy code recipe: default.v1 assistant_id

language: xx

pipeline:

  • name: “SpacyNLP” model: “ko_core_news_lg”
  • name: “SpacyTokenizer”
  • name: “SpacyFeaturizer” pooling: “mean”
  • name: DIETClassifier
    epochs: 165 constrain_similarities: True

policies:

  • name: MemoizationPolicy
    max_history: 5
  • name: RulePolicy
  • name: TEDPolicy
    max_history: 5 epochs: 20 constrain_similarities: true
  • name: UnexpecTEDIntentPolicy
    max_history: 5 epochs: 20 My project aims to distinguish only the intent, and then I plan to customize certain intents for the client-side source.

Now, onto my questions:

In preparation for other computers utilizing the Rasa library via pip, I would like to ensure that the minimum specifications are met. Could you provide information regarding the minimum and recommended specifications for the computer?

How many simultaneous questions can the Rasa engine withstand if directed at it?

I am interested in creating a custom tokenizer that combines Whitespace and Spacy tokenizer functionalities. However, I am currently using the library. Could you provide guidance on how to create a custom tokenizer in this scenario?

I appreciate your time and assistance with these inquiries.

Best regards,

2

You’ll find some basic resource requirements here but this is very much dependent upon your pipeline and policies.

There is some load testing info here

For custom tokenizer, take a look at the custom graph components page.

Also note that you should not be installing the rasa-core and rasa-nlu packages. These are no longer updated and the versions you referenced are both old and incompatible with 3.6.2. You only need the rasa package. The core & nlu packages were consolidated into the single page with the 1.0 release several years ago.