DOCSEARCHbyAlgolia
Submit Docs

Dydu Documentation

website·Indexed 35 days ago

https://docs.dydu.ai/

Documentation for the Dydu chatbot platform, covering bot creation, knowledge base management (connaissances), and deployment via Dydubox and Teams. Includes guides on slot filling, decision trees, behavioral targeting, and integrating external content sources and LLM models such as VertexAI Gemini.

Tokens
54.3K
Snippets
64
Records
426
Agent score
53%

What's inside Dydu

  1. Overview of the Dydu conversational bot platform

    Dydu (Do You Dream Up) is a software solution for creating and managing conversational robots (chatbots). It is designed for versatility across various business applications, including customer service, support automation, and internal knowledge sharing. The platform is built for ease of use, allowing non-technical users to design conversations and deploy bots with minimal training. Key capabilities include intuitive knowledge configuration, learning tools, content management, integration options, and detailed performance analytics.

  2. Overview of predefined response templates

    Dydu provides three main response templates to make chatbot interactions more attractive and guided:

    • Quick Reply: Predefined buttons that guide users to a new piece of knowledge or an external URL.
    • Product Sheet (Fiche produit): A visual presentation of a product or service including an image/gif, title, integer, subtitle (description), and up to 3 buttons.
    • Carousel: Similar to a Product Sheet but allows multiple cards to be displayed in a scrollable format.

  3. Overview of Theme Management in Dydu

    Themes represent the primary subjects addressed by your bot and can be organized into a hierarchy of themes and sub-themes. Managing themes provides three main benefits:

    1. Knowledge Classification: Allows for easier retrieval of knowledge elements based on the subject they treat.
    2. Dialogue Analysis: Enables the dialogue engine to calculate analytics based on the subjects discussed.
    3. Dialogue Review: Allows users to filter and display only the dialogues related to a specific theme on the dialogues page.

  4. Overview of Context Conditions

    Context Conditions are used to memorize specific information during a conversation. This allows the bot to provide different responses based on previously gathered data and avoids asking the user the same question multiple times. For example, if a user specifies a product type early in the chat, the bot can store this in a variable and use it to tailor all subsequent answers regarding that product.

  5. Overview of Phrasing Groups (Groupes de formulations)

    Phrasing Groups are sets of words and terms with similar meanings or intentions. They allow developers to group multiple variations of a question (e.g., 'How do I activate my account?' and 'What is the activation procedure?') into a single group, simplifying the knowledge base and improving the bot's understanding without needing to create separate questions for every variation. Groups can be nested (using other phrasing groups inside a parent group) to expand the bot's comprehension capabilities.

  6. Understand the concept of Knowledge in Dydu

    In Dydu, a 'knowledge' (connaissance) is defined as a pair consisting of a question and a response.

    • Question: Represents the end-user's intent that the chatbot must understand.
    • Response: The content configured to be returned to the user.

    Knowledge can be simple (one question, one answer) or structured as a decision tree, allowing the chatbot to ask follow-up questions before providing a final answer. Knowledge entries can be organized by theme to form the chatbot's knowledge base.

  7. Understand Dydu Integration Concepts

    Dydu provides several interaction layers:

    • Chatbot/Chatbox: The conversational assistant interface for natural language text dialogue.
    • Livechat: A service that connects a user directly to a human operator via the chatbox.
    • Meta Bot (Questionnaires): A feature that allows a primary bot to search for and retrieve knowledge from other bots.

  8. Retrieve conversation and user metadata using the dialog object

    The dialog object provides access to various metadata about the current user session and conversation state.
    // Get the current consultation space
var space = dialog.currentConsultationSpace();

// Get the user's language
var language = dialog.language();

// Get the current URL
var url = dialog.userURL();

// Get user identification (e.g., IP address)
var id = dialog.userIdentification();

// Get the user's Operating System
var os = dialog.userOs();

// Get the user's browser
var browser = dialog.userBrowser();

// Get the current user sentence/query
var sentence = dialog.currentUserSentence();

// Get the bot's current answer
var answer = dialog.currentAnswer();

  9. Organize and analyze knowledge base by themes

    The Dydu knowledge base can be organized into themes (thématiques) to create a hierarchy of information. This organization allows users to track statistics on the most frequently addressed or used themes, including the total volume of use and the percentage of distribution for each theme.

    Users can drill down into the specific knowledge entries that compose each theme and export the resulting analytical data to an Excel file via the 'Export vers Excel' button.

  10. Understand quality alert types and triggers

    The knowledge base is continuously monitored by quality verification processes. Alerts are categorized into three levels: Error (highest priority), Warning, and Information.

    Common triggers for quality alerts include:

    • Response Length: Responses that are too long may be cut off in the UI if the user doesn't scroll.
    • Question Length: Questions that are too long may not display fully above the text area.
    • Invalid Hyperlinks: Links to pages that no longer exist.
    • Missing Theme: Knowledge without a theme makes it difficult to determine the user's subject.
    • Cross-Knowledge Links: Verification that links redirecting users to related knowledge are valid.
    • Expiration Dates: Notifications to update, change status, or delete knowledge when an expiration date is reached.
    • Start Dates: Notifications to change knowledge status to 'Published' on a specific start date.
    • Failed Test Conversations: Daily automatic replays of saved test dialogues; an alert is triggered if the dialogue deviates from the saved version.

  11. Visualize knowledge dependencies with the Knowledge Map

    The Knowledge Map (Carte des connaissances) provides a visual overview of your knowledge base, allowing you to visualize links and dependencies between different pieces of information. It is accessed via the path: Contenus > Carte des connaissances.

    The map uses arrows to represent four specific types of modeled links:

    • Redirections
    • Reformulations
    • Bounce conditions (Conditions de rebond)
    • Models (Modèle)

  12. Monitor bot performance using BMS Exploitation reports

    The Exploitation section of the Bot Management System (BMS) provides pre-generated reports to analyze and steer bot performance. These reports are based exclusively on production conversations; test conversations are excluded from all statistical calculations.

    Available metrics include:

    • Visitor and conversation counts
    • Average interactions per conversation
    • Interaction and conversation qualification (measuring the bot's understanding and ability to provide answers)
    • User satisfaction levels regarding provided answers
    • Most frequently discussed themes
    • Most utilized knowledge base entries

    All indicators can be filtered by time period, language, or consultation space.