Genesys Cloud - Developer Community!

 View Only

Sign Up

  • 1.  Accessing the Knowledge Article Confident Scores in Digital Bot Flow

    Posted 11-05-2025 06:11
    Edited by Felix Mensching 11-05-2025 06:11
    Hi,
    I am looking for some support on the handling of knowledge articles and intents.
    I would like to handle the Disambiguation of  Digital Bot Flow myself.

    Hence, I need the information for the confidence levels of intents and knowledge articles.

    For Intents, I can access the confidence scores using the variables Session.LastCollectionConfidence, Session.LastNLUCollectionConfidence
    Is there something similar for Knowledge articles? (Understand confidence levels - Genesys Cloud Resource Center)

    I know that I can access the collection "Knowledge Search Results", but the collection object does not contain any confident scores as stated by @Rory Brennan in another community post https://community.genesys.com/discussion/knowledge-confidence-score
    => the variable "Flow.searchResults[0].confidence" is not available.

    Does someone know how to access this or handle the confident scores of knowledge articles within the flow?

    Thank you for your support!


    #Architect
    #WebMessaging

    ------------------------------
    Felix
    Miele & Cie. KG
    ------------------------------



  • 2.  RE: Accessing the Knowledge Article Confident Scores in Digital Bot Flow

    Posted 11-05-2025 08:55
    Edited by Luiz Rosa 11-05-2025 09:12

    Hi Felix Mensching,

    You can control the knowledge behavior in your digital bot flow using this guide: Control knowledge behavior in your voice or digital bot flow . The Knowledge Search Results collection returns a list of articles most related to the intent, but it doesn't include the confidence score directly.

    Natively, the bot already performs this search, but if you want to add more business rules, as I did, I configured it to return up to the top three articles and ask the user whether the article was positive or negative to collect feedback.

    Exemple:

    Another alternative is to use the API search:
    /api/v2/knowledge/knowledgebases/{knowledgeBaseId}/documents/search
    It also returns the search results.
    Here’s an example:

    Body:

    Response:

    Hope this helps.



    ------------------------------
    Luiz Rosa
    Full stack developer
    ------------------------------



  • 3.  RE: Accessing the Knowledge Article Confident Scores in Digital Bot Flow

    Posted 11-12-2025 10:15

    Hi Luiz,

    thank you very much for the answer on this.

    So basically, we cannot get sth. like the confidence score from knowledge articles search.

    But what you show here is that I can invoke the search for knowledge articles myself through a data action and handle the response myself. I also get the categories and labels associated with the result If I see it clearly.

    If I get this right, this would allow us to access the categories of the knowledge articles and cycle through them if we like to suggest some articles of the same category, right?

    Is there a possibility to get knowledge articles by category for example?

    BR, Felix



    ------------------------------
    Felix
    Senior Consultant
    Miele & Cie. KG
    ------------------------------



  • 4.  RE: Accessing the Knowledge Article Confident Scores in Digital Bot Flow

    Posted 11-12-2025 10:35

    Hi Felix Mensching,

    Yes, you can search by category. Here's an example:

    The response will return only search results related to the category.

    So you can know the category ID, use this endpoint:

    /api/v2/knowledge/knowledgebases/{knowledgeBaseId}/categories



    ------------------------------
    Luiz Rosa
    Full stack developer
    ------------------------------



  • 5.  RE: Accessing the Knowledge Article Confident Scores in Digital Bot Flow

    Posted 18 days ago

    Yes, it's possible to measure knowledge base performance and confidence levels, even though NLU intent variables don't exist for knowledge articles inside bot flows.

    For knowledge articles, confidence can be inferred and measured indirectly from interaction feedback signals collected after the article is shown to users in the bot conversation.

    Within the platform, end-user feedback (thumbs up/down, article usefulness, etc.) provided immediately after a knowledge article is displayed becomes a key performance indicator for how confidently the knowledge base is performing.

    You can then use built-in optimization capabilities such as:

    • Knowledge base analytics and performance dashboards

    • Optimizer and tuning recommendations

    • Feedback trend tracking for articles

    With these, you can assess system-generated confidence guidance and fine-tune disambiguation or intent resolution logic manually in your digital bot flows.

    So while there is no direct Flow.searchResults[0].confidence variable, the confidence layer exists in practice through measurable feedback + optimizer insights, which can be actively monitored and improved using native Genesys tools.

    Hope this helps! 🚀



    ------------------------------
    Mateus Nunes
    Pre sales
    ------------------------------