> ## Documentation Index > Fetch the complete documentation index at: https://koreai.mintlify.app/llms.txt > Use this file to discover all available pages before exploring further. # Conversation Management This document covers the core conversation management features of the AI Agent: standard responses, interruption handling, small talk, and alert tasks. | Feature | Description | | --------------------------------------------- | -------------------------------------------------------------------------------------------- | | [Standard Responses](#standard-responses) | Pre-defined NLP interpreter messages triggered by events, user inputs, or system conditions. | | [Manage Interruptions](#manage-interruptions) | Configure how the AI Agent handles mid-task interruptions such as FAQs or small talk. | | [Small Talk](#small-talk) | Enable casual, non-task conversations to make interactions feel more natural. | | [Alert Tasks](#alert-tasks) | Set up proactive notifications the AI Agent sends based on triggers or schedules. | ## Standard Responses Standard responses are pre-defined NLP interpreter messages the AI Agent displays based on events, user inputs, or system conditions. All defaults can be customized, and channel-specific overrides can be added. Go to **Automation AI > Conversation Intelligence > Conversation Management > Standard Responses**. For each category, the trigger condition is listed alongside the default response(s). The NLP interpreter randomly selects from multiple responses at runtime. ### Response Categories | Category | Description | | --------------------- | ---------------------------------------------------------------------------------------------------- | | **Statements** | Responses triggered by AI Agent or user actions — for example, task cancellation or invalid input. | | **Queries** | Questions the AI Agent asks for more information — for example, *Which field do you want to change?* | | **Errors & Warnings** | Messages displayed during errors — for example, *End date cannot be before the start date.* | | **Questions** | General questions posed to the user that don't fit other categories. | | **Choices** | Yes/No prompts — for example, *Would you like to receive push notifications?* | | **Greeting** | Initial responses when the app starts or when a user returns after inactivity. | ### Statements Responses triggered by actions from the AI Agent or user. | Message | When It Appears | | --------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | When the user confirms to close the conversation in a language and switch to another language | User ends a conversation in one language and starts the next in a different language. | | When only field prompt is present in the OOB | Internal field prompt for action and alert tasks. Not recommended for customization by enterprises. | | When language preference is set | User switches language using *cheat lang* or picks from language options. | | When task is not accessible to user | Developer tries to access an app task they don't have permission for. | | User confirms alert setup | An alert is set up successfully (Alert Task only). | | User discards/abandons the current activity | User says *discard*, or the task fails due to max retries, invalid LoV, or cURL error. | | User implies to end the conversation | User says *bye*, *good night*, etc. Shown when the Follow-up Intents list is presented at dialog end. | | Location provided by user is invalid | AI Agent cannot validate input for a Location-type entity. | | Amending an Alert field | User says *Yes* to amend alert parameters (Alert Task only). | | Amending a task which has no fields | Task has no entities to amend (Alert, Action, and Information Tasks). | | Ask at least one optional field | User did not provide any optional field value when at least one is required (Action Task only). | | Topic retry limit exceeded | Developer mode only. Dialog looped through the same nodes beyond the allowed limit. | | Universal fallback response | AI Agent cannot generate a response for any reason, or maximum node execution limit exceeded. | | Server returned authorization failure | Authorization failure occurred during script or API execution. | | User selects a valid option from a list | User selects an action/alert/info task from the listed options. | | User selects an action | User selects an action task; AI Agent confirms and asks for required parameters. | | Alert is identified, ask for alert time | Displayed when Trigger Interval is set to *At a specific time*. | | Alert is identified, ask for alert time interval | Displayed when Trigger Interval is set to *Every X minutes*. | | User selects an alert | User selects an alert task; AI Agent confirms and requests required parameters. | | Task discarded | User types *discard* mid-task. | | App choosing the default account | Shown during alert amend when AI Agent cannot parse the user's response. | | Confirmation Dialog for Task consolidated | AI Agent confirms captured parameters before executing an Action/Info task. | | Confirmation Dialog for Alert consolidated | AI Agent confirms all details before setting up an alert (if confirmation is enabled). | | Confirmation Dialog for Alert consolidated without fields | Alert has no fields; AI Agent confirms before execution (if confirmation is enabled). | | The alerts will be muted | Developer has muted alert notifications. | | Confirmation Dialog when alerts are muted | Shown when *Mute alert notifications* is enabled for an RSS alert. | | Alert frequency description | Confirmation of alert query frequency for web-service alert tasks. | | Not understanding the input at confirmation | User's response during alert frequency confirmation does not match the expected input. | | No results found for the search string | Typeahead field returned no results. | | Successful task setup | Alert task set up successfully. | | I am not sure what you mean | One or more low-confidence intents found. Usually shown as FAQ suggestions. | | Ambiguity in intent recognition | AI Agent failed to identify user intent; prompts user to rephrase. | | User has selected daily alert — ask for frequency or specific time | Both *frequency interval* and *specific time* options are available. | | App needs to present filter options | Multiple filters available during alert setup. | | App needs to present filter options with exit option | Same as above, but includes an option to skip. | | User finished configuring filters | Filter configuration for an alert task completed. | | User asks *Who am I* | User types *Who am I*. | | User asks for help | User types *Help*; AI Agent lists available tasks. | | Skipped values | Part of confirmation message for Action/Info tasks — shows skipped (optional) parameters. | | Ask user to change dependent field | A dropdown depends on another field, but the Platform cannot fetch data from that field. | | When there is no data | User chooses to change account but responds *no* to the AI Agent's confirmation. | | When date given in invalid format | User enters an unrecognizable date format. Valid formats: YYYY-MM-DD, DD-MM-YYYY, 1 Jan 1982, *Next Friday*. | | Follow-up sentence post task discard | AI Agent confirms task was discarded and offers to restart. | | Confirmation Dialog for task with skipped fields | Confirmation before executing an Action/Info task when some optional fields were not provided. | | Listing all tasks | Displayed when user types *Help* and no task matches. | | Listing all alerts | Displayed when user types *Help* and the app has only alert tasks. | | Listing all FAQs | Displayed when user types *Help* — includes Knowledge Graph collections. | | Listing all alerts after tasks | Shown when app only has alert tasks and intent is not identified. | | Listing all alerts after tasks and FAQs | Shown when all task types are configured and user types *Help*. | | Lists all actions and FAQs | No matching task found; lists actions and FAQs. | | Lists all alerts and FAQs | No matching task found; lists alerts and FAQs. | | When app has no tasks | Developer tries to run an app without any tasks. | | When a server call fails | Webhook or service node call failed; task is discarded. | | User prompt when a task is discarded due to server call failure | API call did not complete successfully. | | Specific app is not available | App is not yet published. | | Field title and value in confirmation | Part of the confirmation step — shows captured field values for alert, action, or info tasks. | | Recursive dialog found | AI Agent recursively called the same dialog and exceeded the limit (100). | | Loop limit reached for a node | Same node invoked repeatedly beyond the allowed number of times. | | Message component doesn't have a message | Message or Service node returned empty content. | | User says thanks | User thanks the AI Agent. | | User says OK, fine, yes | Shown after task completion when the user acknowledges positively. | | User says no, nope | Small talk response to negative acknowledgments. | | User says great, awesome | Shown after task completion when the user responds positively. | | User says that helped, that was useful | Shown after positive feedback (great, cool, fair enough, that helped). | | User says I am done | Shown when user indicates they are finished (done, that's all, that's it). | | Successfully added filter | A filter was added to an alert. | | After successfully adding a filter, showing filter options again | Shown when a filter is applied to an alert (Alert Task only). | | User gave invalid input for filter | Filter name doesn't match any configured filter. | | Resume paused task | User chooses to resume on-hold task. Shown only when Resume option is *Notify the user with a message that the on-hold task is being resumed*. | | Notify the user that we have canceled the paused task | User chose not to resume the on-hold task during confirmation. | | Notify the user that we have canceled the current task to switch to another | Hold option is *Discard the current task and switch to new*. | | When an interruption is identified but hold is not allowed | New intent detected during task execution; new intent added to follow-up list (current task continues). | | Service call timed out | Third-party API call (webhook or service node) timed out. | | Dialog has no transitions | Dialog task has only an Intent node and no further flow defined. | | Manage account | User says *manage accounts*, *add phone number*, *delete account*, etc. | *** ### Queries Questions the AI Agent asks users for more information. | Message | When It Appears | | ---------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------- | | Ask if user would like to switch to a new language (multiple detected) | Multiple new languages detected mid-conversation; user asked to select preferred language. | | Ask if user would like to switch to a new language (single detected) | A different language detected; user offered option to switch or continue. | | Follow-up sentence post task execution or discard | Task completed or discarded; user offered option to restart. | | Confirmation before discard | User types *Discard* or *Discard All* during task execution (applies to select clients). | | User types *Change* or similar words | User changes their mind during alert filter setup; alert setup is discarded. | | User prompted to enter text to search | Typeahead field prompt (Action, Alert, Info tasks). | | Ask the user to enter alert frequency | Frequency can be interval or specific time (Alert only). | | Ask the user to enter alert interval | Frequency is an interval, e.g., every 4 hours (Alert only). | | Ask the user to enter alert time | Frequency is a specific time, e.g., daily at 9:00 am. | | Alert identified — ask for frequency | Prompted based on schedule options defined by the alert developer. | | Prompting user to enter name for alert instance | User installs alert in Kore messaging rooms. | | User needs to select from a static dropdown | Filter field type is *static dropdown*. | | User needs to select from a search results list | Filter field type is *dynamic dropdown*. | | Request for email address for webhook instructions | Webhook alert setup — developer email needed for configuration instructions. | | Query to get the alert instance description | User prompted to provide a description during alert setup. | | Tell me the alphabet or unique word for the action | Displayed before a list of options when no custom template is used. | | I am not sure what you mean after an invalid choice | User's follow-up doesn't match any of the initially identified ambiguous alerts. | | Asking user to pick an alert | A list of alerts is shown; user prompted to select one. | | Asking user to pick an alert after an error | User selected an invalid option from the alert list. | | When user responded with invalid answer to confirmation dialog | User said they want to amend but the AI Agent didn't understand what — confirmation repeated. | | Typeahead field prompt (first field in task) | First field in task is a typeahead type. | | Ask user what to rename the alert | User chose to change/amend the alert name. | | After user amends a parameter — ask if anything else to change | User amended a parameter; AI Agent asks if further changes are needed. | | No results found for search string — ask to try again | Typeahead field query returned no results. | | Follow-up sentence post discard — Part 2 | At alert confirmation, user said *no* twice. | | Listing ambiguous tasks | User prompted to pick a task from a list; response led to ambiguity. | | Ask if user wants to continue the paused task | Shown after the interrupting task completes (Resume option: *Get confirmation before resuming*). | | Ask if user would like to switch | Hold & Resume is set to *Allow the end user to select the behavior*; AI Agent confirms before switching. | ### Errors & Warnings Messages displayed when errors or validation issues occur. | Message | When It Appears | | ---------------------------------------------------------- | ------------------------------------------------------------------------------------------------------ | | Agent transfer fallback response | Agent unavailable, wait time exceeded, or agent transfer failed. | | User exceeds maximum retries for an entity (task on hold) | User exceeded max retries while another task is on hold. | | Dialog task associated to FAQ is not available | The dialog linked to a FAQ was suspended, deleted, or not assigned to the user. | | User needs to authorize or re-authorize | End-user authorization or re-authorization is required. The authorization URL is appended dynamically. | | User authorization has failed or is yet to authorize | Task requires authorization; AI Agent provides an authorization link. | | User's intent recognized, but authorization fails | Service call needs authorization and no account has been authorized yet. | | User authorization failed more than 3 times | User entered wrong credentials three times in a row. | | User provided value is not unique | User's input causes ambiguity when selecting from a partially matched intent list. | | App did not find any options based on user's choice | Option set is empty, or no account credentials are available. | | Discarding task — no options found for list or lookup | User input not found in service call response (dynamic dropdown or typeahead field). | | Task identified, but server call failed | Kore account for the third-party API expired or failed. | | User enters invalid email | Invalid email address entered for an alert, action, or info task. | | User enters invalid country | Invalid country name entered. | | User enters invalid URL | Invalid URL entered. | | User enters invalid phone number | Invalid phone number entered. | | User enters invalid number | Non-numeric input entered for a number field. | | Invalid date format | Invalid date format entered. | | Missing date | No date provided for a date field. | | Ask user to enter time for a field | Time missing for a DateTime entity in a dialog task. | | Ask user to enter time | User provided only a date for a DateTime entity (e.g., scheduled appointment). | | Invalid time — ask to try again for a field | User's time input for a DateTime entity is missing or invalid. | | Invalid time — ask to try again | User entered an invalid time value (e.g., Aug 29 at 25). | | Missing date/time | Error prompt for DateTime entities in Actions and Alerts. | | You have chosen a wrong choice | Sub-response used when a user selects an invalid option from a list. | | Cannot understand your request | AI Agent fails to identify intent; lists available tasks and FAQs. | | App is selected, user asked for reauthorization | API call failed due to expired credentials. | | Listing all tasks after wrong choice | User selected an invalid option; app has only Actions/Dialogs/Info tasks. | | Listing all alerts after wrong choice | User selected an invalid option; app has only Alert tasks. | | Listing all FAQs after wrong choice | User selected an invalid option; app has only FAQs. | | Listing all FAQs — not understanding request | App has only a Knowledge Graph and fails to understand intent. | | Listing all actions — not understanding request | App has only Action tasks and fails to understand intent. | | Listing all actions and alerts — not understanding request | App has only Action and Alert tasks and fails to understand intent. | | Listing all actions and FAQs — not understanding request | App has only Actions and a Knowledge Graph and fails to understand intent. | | Listing all alerts — not understanding request | App has only Alert tasks and fails to understand intent. | | Listing all alerts and FAQs — not understanding request | App has only Alert tasks and a Knowledge Graph and fails to understand intent. | | Listing all tasks — not understanding request | App fails to understand intent; shows all available tasks. | | Listing all tasks and alerts after wrong choice | User selected an invalid option; app has tasks and alerts. | | Listing all alerts, tasks, and FAQs after wrong choice | User selected an invalid option; app has all task types. | | Lists all actions and FAQs after wrong choice | App has only Actions and a Knowledge Graph; user selected an invalid option. | | Lists all alerts and FAQs after wrong choice | App has only Alerts and a Knowledge Graph; user selected an invalid option. | | Task discarded — user exceeded maximum allowed attempts | User entered an invalid entity format five times. | | Entity component in dialog has no error prompt | Entity node has no error prompt defined. | | Entity component in dialog has no message attribute | Entity node has no message prompt defined. | | User input exceeds maximum sentence or word limit | Input exceeds the allowed character or word count. | ### Questions General questions the AI Agent poses to users. | Message | When It Appears | | ---------------------------------------------- | ------------------------------------------------------------- | | When user input matches and frequency is asked | User input led to ambiguity while setting an alert frequency. | | Ask user to input alert frequency | Prompt for frequency at which the alert task will run. | ### Choices Yes/No prompts used during alert and task setup. | Message | When It Appears | | ----------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------- | | Search results displayed (type-ahead search) | User input triggered a typeahead search in an alert filter (Alert Task only). | | Confirming alert frequency | App confirms current alert trigger frequency with the user. | | Repeat confirmation of default alert frequency | User input doesn't match the expected frequency response. | | Confirmation to set up alert with frequency | Confirm the trigger frequency for an alert. | | Prompting user to set up filters | App asks whether the user wants to configure alert filters. | | Prompting user to set up filters as soon as alert is identified | Same as above — triggered immediately on alert identification. | | Asking user to set up push notifications | User wants to amend the alert notification setting. | | Repeat — asking user to set up push notifications | User's response to notification setting doesn't match yes/no. | | Amending the description | User wants to change the alert description. | | Amending the frequency | User wants to change the alert trigger frequency. | | Amending notification settings | User wants to change whether notifications are enabled. | | Prompting user to choose days of the week | User selecting Weekday/Weekend for trigger frequency. | | Query to choose interval for frequency | User prompted to select a frequency interval. | | Query to send webhook setup instructions via email | App checks if developer wants webhook configuration instructions emailed. | | Do you want to change the alert name? | User wants to amend the alert name. | | Query to choose operator while setting up a filter | User prompted to select an operator for an alert filter. | | Options to pick frequency for alert notification days | All interval options available (Daily, Weekday, Weekend). | | When the user needs to choose from provided choices | User prompted to pick from a dropdown in alerts or actions. | | Confirm if user wishes to reuse a pre-saved context | Prompted when a pre-saved context is available for reuse. | | Show possible intervals for alert schedule | User entered a specific time, but only intervals are valid. | | Show closest frequency for alert schedule | User's input doesn't match available frequencies; nearest option shown. | | Show closest time for alert schedule | User's input doesn't match available times; nearest option shown. | | When changing a field | User selected *No* during alert parameter confirmation. | | Prompting to set up filters after alert identified | User provides input to set up an alert that has filters configured. | | Ask if setup instructions should be sent via email (first prompt) | First prompt in alert setup is to ask about emailing setup instructions. | | While choosing options | Shows options from the Platform for a typeahead entity in an action or alert. | | After selecting task — ask about schedule | Alert selected from list; AI Agent asks how often the user wants to be notified. | | Wrong choice for task schedule | User entered invalid frequency for an alert; AI Agent asks again. | | After selecting task — ask for field value | First field in an action or alert is a dropdown. | | User says *no* for alert setup confirmation | User chose not to proceed with the alert configuration; offered to amend. | | After selecting task — user selects wrong choice in dialog | User entered an invalid choice from a non-LoV list. | | User enters wrong choice for alert frequency | Alert frequency value doesn't match options defined by the developer. | | User enters wrong value for a dropdown | Input doesn't match any value in the List of Values. | | List ambiguous LoV values | Multiple partial matches for a List of Values entity; user asked to select. | | User enters invalid choice for an action | User's response doesn't match the presented task list. | | Ambiguous date | User entered an ambiguous date (e.g., 6/9); both interpretations shown. | | Ambiguous airport | User entered an airport name that matches multiple airports (e.g., London). | | Ambiguous location | Geocode service returned multiple location possibilities. | | More than one account to pick from | User has multiple third-party accounts; asked to select one. | | Intent recognized, but more than one account to pick from | Same as above, when intent is identified. | | Custom lookup returned more than one value | Multiple results returned from a custom list query. | | Chosen list is empty after selecting a valid choice | Dependent dropdown has no options because the parent field value returned empty results. | | User selects an invalid choice (dependent dropdown) | Same as above; user's value for the dependent field is invalid. | | Present follow-up tasks at end of dialog | Shown at end of dialog when Follow-up Task setting is *Yes, ask the user to select a follow-up task*. | | Ask user to select entity to amend (ambiguous entity value) | Multiple entities share the same value during dialog execution; user asked to clarify which to amend. | | Max on-hold quantity exceeded — ask user to cancel a task | User tried to put more tasks on hold than the configured maximum. | | User provides intent when prompted for String/Description entity | User input matched an intent when prompted for a string/description entity; AI Agent asks how to proceed. | ### Add and Edit Responses **To add a new standard response:** 1. Select a **category tab** and hover over the target situation. 2. Click **Add Response**. 3. In the **Add Channel Overrides** dialog: * Select a **channel** (or **Any Channel** for all channels). * For channel-specific responses, select a **task** and optionally a **field**. add channel override 4. In the **Response** editor, choose a mode: * **Simple Mode** — Plain text with optional formatting (bold, italics, hyperlinks, lists). * **Advanced Mode** — JavaScript for custom channel-specific messages. Use `response.message` for the default message string and `response.choices` for the options array. 5. Click **Save**. ### Channel Override Examples **Slack:** ```js theme={null} var message = {}; var attachments = []; var attachment = { 'text': "Exception occurred during API execution", 'color': '#FFCC00', 'attachment_type': 'default' }; attachments.push(attachment); message.attachments = attachments; print(JSON.stringify(message)); ``` **Facebook:** ```js theme={null} var message = { text: response.message }; var buttons = []; for (var i = 0; i < response.choices.length; i++) { var choiceItem = response.choices[i]; if (choiceItem.length > 20) { choiceItem = choiceItem.substring(0, 18) + '..'; } buttons.push({ content_type: 'text', title: choiceItem, payload: choiceItem }); } message.quick_replies = buttons; print(JSON.stringify(message)); ``` **SMS:** ```js theme={null} var message = {}; var indexArray = ['a','b','c','d','e','f','g','h','i','j','k','l','m','n','o','p','q','r','s','t','u','v','w','x','y','z']; var text = response.message + '\n'; for (var i = 0; i < response.choices.length; i++) { text = text.concat(indexArray[i] + ') ' + response.choices[i] + '\n'); } message.text = text; print(JSON.stringify(message)); ``` **Email:** ```js theme={null} var message = {}; var text = response.message + ' '; for (var i = 0; i < response.choices.length; i++) { text = text.concat('' + response.choices[i] + '
'); } message.text = text; print(JSON.stringify(message)); ``` ### Customizing Ambiguity Display with `getChoices()` Use `getChoices()` to retrieve ambiguous intent choices and render them in a custom format (for example, as buttons): ```js theme={null} var info = getChoices(); var message = { "type": "template", "payload": { "template_type": "button", "text": "Did you mean", "subText": "You can now customize response for did you mean", "buttons": [] } }; for (var i = 0; i < info.length; i++) { message.payload.buttons.push({ "type": "postback", "title": info[i], "payload": "payload1" }); } print(JSON.stringify(message)); ``` *** ## Manage Interruptions Users rarely follow a linear conversation path — they may switch topics, ask unrelated questions, or provide a new intent mid-task. Manage Interruptions gives you granular control over how the AI Agent handles these situations. Go to **Automation AI > Conversation Intelligence > Conversation Management > Manage Interruptions**. ### Interruptions Hierarchy Settings at more specific levels override broader defaults. | Level | Precedence | Where to Configure | | --------------- | ---------------- | ------------------------------------------------------------------------------ | | **Node** | Highest | Node **Instance** tab > **Advanced Controls** > **Interruptions Behavior** | | **Dialog Task** | Middle | Dialog Builder > **more options** icon > **Manage Interruptions** | | **App** | Lowest (default) | **Conversation Intelligence > Conversation Management > Manage Interruptions** | > If no node or task-level settings exist, the app-level settings apply. When a user provides input after 15+ minutes of inactivity and the input is within the previous session context, it is treated as an interruption. If the user starts a new utterance, the old context is discarded. ### Allow Interruptions | Option | Behavior | | ----------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------- | | **Hold the current task and resume after the new task completes** | Switches to the new intent immediately. Requires a Resume Option (see below). | | **Discard the current task and switch to new** | Discards the current task and switches. Sends the user a notification message. Customizable via **Manage Response**. | | **Switch to new task without notification and discard current** | Silently discards the current task and switches. No user notification. | | **Continue the current task and add new task to follow-up list** | Stays on the current task and adds the new intent to the `FollowupIntents` array. Customizable via **Manage Response**. | ### Do Not Allow Interruptions Disables interruptions at the app level. Can still be overridden at the task or node level. ### Allow the End User to Select the Behavior Prompts the user to confirm whether to switch tasks. Customize the confirmation message via **Manage Response**. Requires a Resume Option. *** ### Resume Options Defines what happens to the on-hold task after the interrupting task completes. | Option | Behavior | | ------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------- | | **Get confirmation before resuming** | AI Agent asks the user (Yes/No) whether to resume the on-hold task. Message is customizable. | | **Notify the user and resume** | Resumes the on-hold task automatically and notifies the user. No confirmation required. Message is customizable. | | **Resume without any message** | Directly resumes the on-hold task with no notification. | | **Always resume without message if the task ended in a single response** | If the interrupting task completed in a single response, resumes the on-hold task silently regardless of other resume settings. | **Example — Get confirmation before resuming:** ``` User: Can you book me a flight for tomorrow? AI Agent: From which city are you flying? User: Los Angeles AI Agent: Where to? User: By the way, what's the weather forecast for tomorrow? AI Agent: Please enter the city for the forecast. User: Los Angeles AI Agent: Weather Forecast for Los Angeles — March 15, 25°C, Mostly Sunny AI Agent: Should I continue with the task 'Book Flight'? User: Yes AI Agent: Enter the name of the destination airport. ``` *** ### On-Hold Quantity Set the maximum number of tasks that can be held at once. Default: **1**. Range: 1-N. When the limit is reached, new tasks are ignored regardless of interruption settings. Tasks resume in reverse chronological order (most recent first). Limit on-hold tasks to 1-2 to avoid confusion. *** ### Node-Level Customization 1. Open the dialog task and select the node. 2. Go to the **Instance** tab > **Advanced Controls**. 3. Under **Interruptions Behavior**, select **Customize for this node** and configure settings. node level customization ### Dialog-Level Customization 1. Open the dialog task. 2. Click the **more options** icon (top-right of Dialog Builder) > **Manage Interruptions**. 3. Under **Interruptions Behavior**, select **Customize for this task** and configure settings. dialog level manage interruptions *** ### Behavior for FAQs | Option | Behavior | | --------------------------------------------------------- | --------------------------------------------------------------------------------------- | | **Respond to FAQs and resume the on-hold task** (default) | AI Agent answers the FAQ and then resumes the on-hold task. | | **Execute the FAQs using Hold & Resume Settings** | Treats FAQ intents like any other intent and applies configured Hold & Resume settings. | FAQ interruption behavior ### Behavior for Ambiguous Intents | Option | Behavior | | ----------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------ | | **Present ambiguous intents to the end-user and proceed as per Hold and Resume settings** | User selects a task; interruption settings are applied based on the selection. | | **Do not present ambiguous intents and continue on the current task** (default) | AI Agent ignores ambiguous intents and continues the current task. | ambiguous interruption behavior ### Behavior for Small Talk | Option | Behavior | | --------------------------------------------------------------- | ------------------------------------------------------------------------------------- | | **Respond to Small Talk and resume the on-hold task** (default) | AI Agent responds to Small Talk by putting the current task on hold, then resumes it. | | **Execute Small Talk using Hold & Resume settings** | Applies configured Hold & Resume settings to Small Talk. | small talk interruption behavior ### Behavior for User Authorization When a user provides unexpected input at an authorization prompt: 1. If **Small Talk** is detected: Small Talk response is shown, and the authorization prompt is repeated. 2. The Platform checks for intents regardless of interruption settings: * **Single intent detected** — Asks the user to confirm discarding the current task and triggering the new one. * **Multiple intents detected** — Shows an ambiguity dialog with an option to ignore and continue the current task. 3. If **no intent** is found — User is re-prompted with the authorization link. *** ## Small Talk Small Talk enables the AI Agent to engage users in casual conversation. AI Agents that respond to social inputs like *How are you?* create a positive impression and improve engagement. Go to **Automation AI > Use Cases > Small Talk**. Intents and FAQs take precedence over Small Talk. If both are detected, the intent or FAQ is handled first. Small talk is language-specific — configure it separately for each enabled language. ### Features | Feature | Description | | ------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Custom Small Talk** | Build a personality and train the AI Agent for casual conversations around your areas of interest. | | **Nested conversations** | Support follow-up queries during a conversation for more engaging interactions. | | **Emoji identification** | The Platform recognizes emojis in user utterances and responds accordingly. | | **Intent-aware** | Interjections like *hello* at the start of an utterance are not treated as small talk if an intent follows — for example, *Hello, I need to book a flight* triggers the book-a-flight intent, not Small Talk. | ### Default Small Talk (Greetings Group) When you create an app, a **Greetings** group is generated automatically. The following messages are included: | User Query | When It Appears | | ---------------------------------------------------------------- | ------------------------------------------------------------------------ | | User says *Hi*, *Hello*, *Hey*, etc. | Shown when the user greets or enters the app name. | | *How are you?* | Shown for greetings like *how are you*, *how's it going*, *wassup*, etc. | | *Who are you?* | Shown when the user asks who the AI Agent is. | | User says *great*, *awesome*, etc. (no task in context) | Shown after task completion when the user responds positively. | | User says *that helped*, *that was useful*, etc. | Shown after positive feedback. | | User says *no*, *nope*, etc. (no task in context) | Small talk response to negative acknowledgments. | | User says *I am done*, *That's it for now*, etc. | Shown when the user indicates they're finished. | | User implies to end the conversation (*bye*, *good night*, etc.) | Shown when Follow-up Intents are presented at dialog end. | | User says *thanks* | Shown when the user thanks the AI Agent. | | User says *OK*, *fine*, *yes*, etc. (no task in context) | Shown after task completion when the user acknowledges positively. | When you open Small Talk in an existing app, you are prompted to migrate Standard Response greetings to Small Talk. This migration is permanent — greetings are removed from Standard Responses and cannot be retrieved if you proceed. ### Set Up Small Talk #### Step 1: Add a Group 1. Go to **Automation AI > Use Cases > Small Talk**. 2. Click **New Group**. 3. Enter a **Group Name** and click **Proceed**. Groups are evaluated in the order they are added. #### Step 2: Add Query-Response Pairs 1. In the Small Talk editor, enter the **User Query** and **Bot Response**. 2. Press **Enter** or click **Add** to save the pair. Additional options per query: * **Alternate questions** — Additional phrasings for the same query. These are pattern-based. * **Alternate responses** — Multiple responses; the Platform picks one at random at runtime. * **Child queries** — Follow-up questions available only after the parent is answered. Up to 3 levels of nesting are supported. Use the drag handle to reorder queries. Add specific queries before generic ones (for example, *how is your day today?* before *how are you*). #### Step 3: Configure Channel-Specific Responses (Optional) 1. Hover over a response and click the **Settings** (gear) icon. 2. In the **Manage Response** window: * Add a default message for all channels. * Click **+ Add Response > New Prompt Message** for channel-specific responses. * Select a **Channel**, then enter a response in **Simple** or **Advanced (JavaScript)** mode. 3. Click **Save**. #### Import Small Talk To import an existing Small Talk dataset: 1. Click the **more** (ellipsis) icon in the Small Talk view. 2. Select **Import**. 3. Download the sample file if needed to verify formatting (JSON or TSV). 4. Select your file and click **Next**. Importing replaces all existing Small Talk data. *** ### Small Talk Context Mark sections of user input patterns with `_` to capture contextual data. This data is stored in `context.smallTalk.matchData` and can be used to personalize responses. **Example pattern:** `I am from _~location Where are you from?` ```js theme={null} // context.smallTalk { "pattern": "I am from _~location Where are you from?", "matchData": { "_0": "chicago" }, "previousMatchData": [] } ``` Multiple tokens are positional (`_0`, `_1`, `_2`, etc.) and may not be contiguous if an optional token is absent. **Example pattern with optional token:** `how is the _[ climate temperature ] { at _~location } ~time` * User says *"how is the temperature in London today"* → `_0: temperature`, `_1: London`, `_2: today` * User says *"how is the climate today"* → `_0: climate`, `_2: today` (`_1` is absent) Access parent-level context from child nodes via: ```js theme={null} context.smallTalk.previousMatchData[i].matchData._0 ``` **Context persistence:** The context is cleared when: * The session ends. * A new primary Small Talk intent is detected. * Any non-Small Talk intent is triggered. *** ### Training After making changes to Small Talk, click **Train** in the Small Talk editor to retrain the assistant. Importing a Small Talk file triggers training automatically. Key rules: * Only 3 levels of child nodes are allowed. * Duplicate top-level questions across groups are not allowed. * Questions are evaluated in the order added within a group; groups are evaluated in the order created. *** ## Alert Tasks Alert Tasks proactively deliver information to users by polling a service on a real-time or scheduled basis. Users can configure alerts to trigger at specific times or when defined conditions are met. Go to **Automation AI > Use Cases > Alert Tasks**. ### Example Use Cases * Time- or location-based promotional offers * Scheduled delivery updates * Inventory level changes * Critical issue notifications * Weather alerts * Cross-system status changes *** ### Create an Alert Task 1. Open the AI Agent. 2. Navigate to **Automation > Use Cases > Alert Tasks**. 3. Click **Create Alert Task**. 4. Configure the **General Settings**: | Field | Description | | ----------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | **Task Name** | The name displayed in the Platform. | | **Display Name** | The name displayed in the Store. Defaults to Task Name. | | **Connection Type** | **Webhook** — app pushes data to a Kore.ai endpoint. **RSS** — Kore.ai polls an RSS feed. **Webservice** — Kore.ai polls a web service using end-user credentials. | | **Connection Mode** | For Webservice only: **REST** or **SOAP**. | | **Input Preconditions** | Intent is detected only if specified context tags are present. Available via **More Options**. | | **Context Output** | Context tags set when this intent is initiated. Available via **More Options**. | #### Advanced Settings | Field | Description | | ------------------------------------- | ------------------------------------------------------------------------------------- | | **Task Install Confirmation Message** | Message shown when the task is successfully set up. | | **Turn Off Confirmation Messages** | **Yes** — skips NLP execution confirmation. **No** — prompts the user before running. | | **Search Keywords** | Keywords users can use to find this task in the marketplace. | | **Task Demo Link** | Link to display a task demo icon in the Store. | ### Authorization Define how end-users are authenticated to receive alert notifications. Not required for Webhook connection types. 1. On the **API Request** tab, expand the **Authorization** section. 2. Select an existing **Authorization Provider** or click **Create New**. 3. Select an **Authorization Type**: | Type | Description | | -------------------------------- | ------------------------------------------------------------------------- | | **Basic Auth** | Standard username/password collection over SSL. | | **OAuth v2 Password Grant Type** | Custom authorization for non-standard web service auth flows. | | **OAuth v1** | Protected resource access via API without exposing user credentials. | | **OAuth v2** | Modern OAuth with specific authorization flows for web apps and services. | | **API Key** | Token-based identification/authentication for incoming app requests. | ### API Requests Configure the API request that fetches data for the alert. The configuration depends on the Connection Type: | Connection Type | Configuration Guide | | --------------- | ------------------------------------------------------------------------------------ | | **Webhook** | Configure a webhook endpoint to receive pushed data. | | **REST** | Define REST API parameters, initializers, pre/post-processors, and response samples. | | **SOAP** | Define SOAP API parameters and the request envelope. | | **RSS** | Define the RSS feed URL and polling parameters. | You can define API Parameters, an optional API task Initializer, API Request pre/post-processors, and expected Response Samples. *** ### Bot Responses After configuring the API request, define how alert results are displayed to the user. | Component | Description | | ------------------------------ | --------------------------------------------------------------------------------------------------------------------- | | **Report Template Definition** | Displays payload data as a table, contact card, or formatted report. Not applicable to Webhook, Action, or RSS tasks. | | **Message Definition** | Defines the look and feel of the alert message delivered to the end-user based on the webhook payload. | ### Alert Settings Define when and how the alert fires. Not applicable to Webhook tasks. | Setting | Description | | --------------------- | ------------------------------------------------------------------------------------------------------------------------ | | **Filters** | Limit alert results by criteria (name, account, amount, text match, etc.). Only matching results trigger a notification. | | **Schedule** | Set polling frequency and schedule options presented to users during setup. | | **Advanced Settings** | Configure the initial alert message set, alert reminders, and duplicate message prevention. | *** ### Version Namespaces Associate Variable Namespaces to scope which variables are used during task execution. This option is only available when Namespaces are enabled. *** ### View Logs View logged events for published alert tasks to troubleshoot request/response behavior. 1. In **Alert Tasks**, click the **Settings** (gear) icon for the task. 2. Click **View Logs**. Each log entry is a JSON object containing the request, response headers and body, and execution time. ```js theme={null} { "requestLogs": [{ "name": "Request 1", "request": { "endpoint": "https://www.example.com/rss/{channel}.xml", "pathFields": { "channel": "latestnews", "lastRun": "2025-10-09T10:20:45Z", "now": "2025-10-09T10:26:37Z", "today": "2025-10-09" }, "payloadFields": {}, "headerFields": { "If-Modified-Since": "Thu, 09 Oct 2025 10:00:00 GMT", "If-None-Match": "\"example-123456\"" } }, "response": { "status": 200, "headers": { ... }, "body": "Example.com Latest News\n..." }, "executionTime": "4860 ms" }], "alertsData": [{ "uxmap": { "channel": "default", "body": "<% var newLink = title; print(newLink); %>" }, "payload": "{ ... }", "response": { "channel": "default", "body": "Example.com launches new AI-powered feature\n" } }] } ``` Refresh the **Logs** page to view the latest entries. ***