Chat
Chat API
API Change log
Released 2024.12.10
-
New feature: We now support ratings for intents from Knowledge and Search federated bot sources. As part of our initial delivery, we have released the logic that allows these kinds of ratings to be recorded in the Chat API. In future iterations, you will be able to manage this feature's activation through the Inbenta Chat administration console.
-
As part of an effort to improve comprehension and add clarity across our products, we have renamed "Chatbot" to "Chat." When using Chat, "Knowledge Base" will now be called "Intent Base."
Released 2024.11.12
-
Fixed: Chatbot applies several discard filters before returning results for a user query. However, some of them were resulting in different final results depending on the Knowledge Base version. This is now fixed.
-
New feature: We now avoid displaying intent titles for those intents that have been configured as not matchable by natural language matching (NLM), so internal knowledge base information is not displayed to the end user.
Released 2024.10.01
-
New feature: When returning a generative AI answer in a conversation, the Chatbot will include information about each content used to generate the answer, including the title, and optionally the URL or summary of the content. The Botmaster will also be able to check the contextual information in Dashboards.
Released 2024.09.17
-
The API now allows users to set variable values when the Chatbot accesses specific intents, so the dialog workflow can be controlled without using webhooks or custom adapters.
Released 2024.08.20
Released 2024.08.06
-
Fixed: When setting a variable with a webhook, blank spaces were not removed if they were at the beginning or the end of the value. This caused issues when selecting the value in a form. This is now fixed.
Released 2024.07.23
-
Added: The Chatbot API will no longer skip the intent answer text, if the intent has a webhook with no form fields.
-
Fixed: The API response format was different when the user exceeded the character limit in the input. Now this is fixed, and the response format is consistent for all use cases.
Released 2024.06.26
Released 2024.03.05
-
Fixed: When using the
POST /inbtrck/externalMatching
endpoint to track external matchings and their corresponding impressions, they were not registered correctly if the Chatbot returned the no-results dialog. This is now fixed.
Released 2024.01.09
-
We added the new endpoint
POST /externalMatching/userQuestion
. It allows you to track external user questions while generating a log ID that can be used to track external matchings.
Released 2023.05.16
-
New feature: When using variables with the data type “Date”, the Chatbot tried to capture date values with the MM/DD/YYYY format, or, if that failed, the DD/MM/YYYY format. Since this behavior could lead to errors, we added the possibility to define the date format for your instance in Settings > General Settings under Conversation Settings > Regional Settings. Users can now choose between three formats: Auto (current behavior), MM/DD/YYYY or DD/MM/YYYY.
Released 2023.05.03
-
New feature: We improved the tracking for Federated Bots. Users can now configure a tracking source name for each Federated Bot source and use it to filter the Dashboard. In addition, Federated Bot results are now also tracked in the external instance and all interactions from a single Chatbot conversation are tracked in the same session in that instance.
-
Improvement: The WEBHOOK_UNEXPECTED_ERROR
tracking event will now allow to check if the error was produced by the webhook or by our webhook request service. When the error is produced by our webhook request service, “chatbot_webhook_client_error” will be stored inside the WEBHOOK_UNEXPECTED_ERROR
tracking event data.
Released 2023.03.21
-
New feature: We created a new detection scope for variables called “Persistent data field”. This detection scope allows the Chatbot to capture the variable value after launching the data field prompt question, but the variable value persists during the conversation, as it only resets when at conversation start.
-
Improvement: From now on, when a user clicks on a result below the minimum score, offered as a post-response behavior, the
CLICK_BELOW_MIN_SCORE
data key will be tracked, to display the event in Dashboard > Sessions > Details.
-
Fixed: The end transaction messages defined in the Intent Editor were not displayed when an intent was returned after clicking on it in a list of multiple results. This is now fixed.
Released 2023.03.07
-
Fixed: In Chatbot conversations, if an intent was redirecting to another inactive intent, a conversation error message was returned to the user. This is now fixed and we no longer show the error message to the user, but simply register it in the report in Dashboard > Sessions > Details.
Released 2023.02.21
-
New feature: Users can now create dialog transitions based on more than one condition. You can for example use this feature to transition a dialog to a specific node when the variable ACCOUNT_TYPE has the value “savings” and the variable USER_AGE has the value “< 21”, to provide information about savings accounts for young people. For more information, see the Help Center article Creating variable conditions in the Dialog Manager.
Released 2023.02.07
-
Fixed: Some user inputs were getting too many possible results which led to timeouts. We reduced the number of considered results to fix this.
-
Fixed: Folder filtering within the Chatbot API using the Editing Knowledge Base version was not working as expected. Intents located in lower level sub-folders were not being returned. This is now fixed.
-
New feature: We now use detection rules to detect Yes/No answer. If this detection fails for some reason, then the old detection based on Local Grammars is used.
Released 2023.01.24
-
Fixed: Action parameters defined as “Input” and “Output” were not working properly in the Editing environment, and therefore webhooks were not receiving those action parameters correctly. This is now fixed.
Released 2022.10.18
-
Fixed: Surveys can now be submitted through the
POST /surveys/submit
endpoint even if the Chatbot conversation session has expired.
-
Fixed: Changes to related intents could not be tested in Editing. This is fixed now, and you can test those changes before publishing to Test or Production by selecting the "Editing" environment in Test & Linguistic Tuning > My Chatbot or Chatbot Debug.
-
Added: The Chatbot API now allows sending and logging conversion events.
-
Fixed: The intent performance data for re-visitable intents returned in a list of multiple results was not being tracked properly. This is now fixed.
-
Fixed: When a user asked a question after seeing a dialog node with expired child nodes, clicks were tracked for those child nodes, even though they were not displayed. This is now fixed and the user questions are tracked correctly.
Released 2022.10.04
-
Fixed: When a popular intent had been deleted, the Chatbot was not displaying the most viewed intents as post-response behavior in the Editing environment. This is fixed now, and the most-viewed intents are returned without the deleted intents.
Released 2022.09.20
-
Fixed: Popular contents are no longer retrieved when non-default user types are set.
-
Fixed: An error no longer occurs when the Chatbot tries to capture a variable value using a Boolean value.
-
Fixed: The Chatbot API now requires much less memory to handle dialogs and their potential pathways, which prevents it from reaching its memory limit when handling large dialogs with many different possible routes.
Released 2022.09.06
-
Fixed: Additional clicks on revisitable related intents are no longer tracked twice.
-
Fixed: We updated the text used in “multiple-options” direct answers to remove any mention of the “click” action. It now says "There is more than one answer which may help you find the information you need. Please select one of the highlighted answers below for more detail:"
-
Improved: The
POST /inbtrck/events
endpoint no longer tracks internal Chatbot events. This will prevent data inconsistencies in session tracking data.
-
Improved: An action-field-failed flag is now returned when the Chatbot is not able to capture a variable value in an action field. This way, developers can know when this occurs and take action if necessary.
Released 2022.08.23
-
The Chatbot no longer logs conversations containing a Dynamic Redirect as separate intents. Now, all redirections are considered part of the same intent.
-
Fixed: We improved performance on retrieving popular contents in the “Editing” environment.
-
Fixed: We changed the post-response behavior to only include intents with matching limited by variable conditions in the most viewed results if all of the corresponding variable conditions are met.
Released 2022.08.09
-
The Chatbot API now allows to track results from external APIs.
-
The Chatbot API now tracks internal events, which allows users to know when a Chatbot dialog is started. (This will be used in the Dialog Performance Data feature which will be released soon.)
-
Fixed: The Chatbot was unable to continue the conversation after trying to add the previous answer message to the Post-response answer while there was no previous answer. This is now fixed.
Released 2022.07.26
-
We reduced the Chatbot’s response time by a few milliseconds in tests using the "Editing" Knowledge Base version.
-
Variables can be inserted within the Answer of an intent so that their captured value is displayed in the Chatbot SDK. However, if a value contained special characters, the Chatbot API did not capture it as expected, and the value was not correctly displayed in the SDK. This is fixed now.
Released 2022.07.12
-
We optimized the internal API memory used for bot conversation history in order to store less data and make the Chatbot API faster.
-
Until now, all user input was stored in conversation history even if it had invalid information. Now we are only storing valid information.
-
When a list of most-viewed intents is returned as post-response behavior defined in the Extended Intent Attributes section of the Intent Editor, intents within a dialog that are set to be presented as a result of automatic transitions are no longer returned in that list. This helps ensure that end users see the most relevant most-viewed intents only.
Released 2022.06.28
-
The Chatbot now skips the last check question in dialogs with the transition type "Offer yes/no questions" when the last check question is empty.
-
Fixed: When the Chatbot was not able to obtain a variable value from users answering a data field prompt in nested intents, and therefore offered multiple options as revisitable links, it stopped working properly. This is now fixed.
Released 2022.06.14
-
Fixed: Intents and child nodes with the “Considered for most-viewed results” switcher deactivated were returned as most-viewed results in tests using the Editing KB Version. This is now fixed.
-
We reduced the Chatbot’s response time by a few milliseconds when using the Editing Knowledge Base version.
-
The Chatbot now returns all post-response options together with the last answer. As a result, in the SDK and test tools like the Chatbot Debug and My Chatbot, the last answer and the post-response options will all appear in the same bubble.
-
When the Chatbot returns the most-viewed intents as post-response behavior using the Editing KB version, scheduled and expired intents are no longer included in the list.
Released 2022.05.31
-
We improved the Chatbot API to allow webhooks to reset variable values. When set to NULL, if the variable has a default value defined, the variable will have the default value, and if no default is defined, the value will be NULL.
-
We slightly reduced the Chatbot’s response time in the Editing Knowledge Base version.
-
The Chatbot API now uses the following aliases for tracking routes so that ad blockers do not automatically block requests: POST /inbtrck/session/user
for POST /tracking/session/user
and POST /inbtrck/events
for POST /tracking/events
. The original endpoints remain active but are now marked as deprecated in the Developers Portal documentation, and the Chatbot SDK will start to use the new routes.
-
Fixed: When used with the Editing Knowledge Base version, the Chatbot API was returning intents that were scheduled for future activation or that had already expired. This is now fixed.
Released 2022.05.17
-
Fixed: When the Chatbot returned multiple results and one of them was a revisitable link that contained an information request and transaction, if the Chatbot nested an intent during one of the information requests, it then returned an error message and was unable to continue the conversation.
Released 2022.04.20
-
New Feature: The “Editing” Knowledge Base version in the Chatbot Debug now includes full support for list variable detection. Whereas previously only exact matches between user input and list values were captured in this environment, now matches using Inbenta NLP technology are captured as well. This feature will begin to take effect on April 20th and we anticipate that it will be activated in all Chatbot instances within a week.
-
New Feature: The Chatbot is now prepared to offer an alternative to the no results response by either offering additional results that match the user’s intent but are below the minimum score configured in Settings > General Settings, or by suggesting some of the most often viewed results from the Knowledge Base.
-
The Chatbot can be configured to search the Knowledge Base for results matching the end user’s reply to the first field of an information request when it is not able to capture a variable value. For this behavior to happen, the information request must be within a dialog child node transitioned to manually or occur after the user has already answered at least one other data field prompt in the dialog.
Released 2022.04.05
-
To facilitate voice recognition, the Chatbot API now returns the ‘dataType’ of the variable requested in the ‘actionField’ within the POST /conversation/message
response. For more information on the contents of ‘actionField’ visit Chatbot API Routes and API Answer Types.
-
Fixed: The Chatbot was unable to continue a conversation after trying to return a dialog node with an expired scheduled deactivation date. Now, the Chatbot skips any nodes that are inactive.
Released 2022.03.22
-
The onDisplayChatbotMessage(messageData)
subscription now returns the folder IDs of the intent within the additionalData
attribute.
-
We added information to the MAX_EXECUTION_TIME_EXCEEDED_ERROR tracking event to better understand how some Chatbots are creating infinite loops that eventually trigger a timeout error. Additionally, the request ID and event order in the request are stored in all events sent to Dashboards.
Released 2022.03.08
-
When the user answers a data field and the Chatbot is not able to capture a value, it performs a search. If the Chatbot finds an intent matching the answer, it returns that intent instead of an error message. Until now, when the Chatbot was able to return an intent, an error message still was tracked by the Chatbot API. It is not matched anymore.
Released 2022.02.08
-
The Chatbot API now returns multiple results retrieved when searching the knowledge base upon failing to capture the user’s response to a data field prompt. The clicked intent is nested and then the Chatbot returns to the previous intent.
-
Being in Editing version, when a non-identifying title is matched in a subset search, it is now returned as the intent title. The intent’s original title is returned within inbMainTitle
as well.
-
Fixed: When a dialog had both nodes with automatic transition and nodes with manual transition, users could revisit dialog node buttons. In order to avoid possible issues, dialog buttons are not revisitable anymore when they are preceded by an automatic transition.
-
Fixed: When using POST /conversation/message
, if the payload contained an empty message and a directCall
with a value, Chatbot API prioritized the message parameter instead of the directCall
. This is now fixed, and directCall
is always prioritized if sent.
Released 2022.01.25
Released 2022.01.12
-
Feature: The Chatbot API is now prepared to begin an information request within an intent, receive and respond to a second intent, and then return to and complete the initial intent. This concept, called nested intents, involves searching the knowledge base for results upon failing to capture the user’s response to a data field prompt as a value for the variable associated with that field.
-
The endpoint GET /surveys/{surveyID}
now returns whether or not the survey is active.
-
We deprecated the allowUserToAbandonForm
parameter in the /conversation/start
route. This means the Chatbot will always behave as if this parameter was set to its default value of true, and users will be asked if they wish to continue a form or information request if they exceed the set number of data field retries.
Released 2021.12.14
-
The Chatbot API now tracks the Chatbot’s language in the CONVERSATION_START_SETTINGS event.
-
For improved security, the Chatbot API now sends a generic User-Agent header value on all webhook requests with the value “Inbenta Chatbot API v1”.
Released 2021.11.16
-
The Chatbot API now stores in caché the variables and attributes definition, and the categories hierarchy. It also preloads the welcome message in order to improve the performance of the Chatbot Debug using the Editing knowledge base version.
-
Added: We now allow the variable capture when the conversation starts.
Released 2021.09.07
Released 2021.08.24
-
The Chatbot API will now register an error if a webhook response takes longer than the 30-second timeout.
-
Fixed: The Chatbot API was not returning the displayed title if it was not set as the identifier. This is now fixed and we return it when the query matches it by natural language.
Released 2021.08.10
-
We improved the security of HTTP headers.
-
Added: We now track in the Chatbot session any webhook that does not pass the Chatbot API security validation (i.e. webhooks using untrusted URLs).
Released 2021.06.15
-
Fixed: The
GET /surveys/{surveyId}/start
was not returning the survey page id.
Released 2021.05.04
-
Fixed: We improved the validation of conversation message inputs.
Released 2021.04.06
-
Added: New
GET /surveys/{surveyId}/start
and POST /surveys/{surveyId}/submit
endpoints that allow to programatically answer a survey using the Chatbot API.
Released 2021.03.23
Released 2021.02.09
-
Updated: We ensured that debug mode cannot be enabled in any case in Production.
Released 2021.01.26
-
Improved: We improved the RTT performance upon starting conversations.
-
Fixed: Multiple attribute groups in RTT are now formatted correctly.
Released 2021.01.12
-
Added: "Skip last question" is now split into two different settings for multiple and dialogs
-
Added: Performance improvements for RTT (cached actions and forms for direct calls)
-
Added: We return titles of matched contents (debug mode).
-
Improved: We improved webhook error feedback.
-
Fixed: The min score filter is applied as the first filter in content results.
Released 2020.12.23
Released 2020.12.09
-
Added: We added the
messageList
attribute in conversation history.
-
Fixed: We improved the performance of dialogs in RTT mode.
Released 2020.11.11
-
Fixed: There was an issue with the encoding of source names with special characters. This is now fixed.
Released 2020.10.28
-
Improved: As part of our continuous maintenance process, we reviewed and cleaned up the code to remove obsolete dependencies.
Released 2020.10.14
Released 2020.09.30
-
Added: The API now sends the source parameter to the surveys module. This makes it possible to track the source of the survey.
Released 2020.09.02
Released 2020.08.19
-
Updated: Previously, the Chatbot API did not allow you to set the Dashboard source when Real Time Testing was enabled but with the debugging mode disabled. This is no longer the case and you can now set the source.
-
Updated: The Multiple options label was updated in German to “Diese Antworten könnten Ihnen helfen:”
Released 2020.07.22
-
Fixed: The API did not properly insert variable values into content attributes in the Real Time Development environment. This is now fixed.
-
Fixed: The API did not properly apply the minimum score in the Real Time Development environment when the score was very low. This is now fixed.
Released 2020.07.01
-
Feature: Return the content condition data for all matched contents in debug mode.
-
Fixed: Results were not being returned in score order in real time development mode. This is now fixed.
-
Fixed: Attribute objects hidden in certain profiles were being returned in real time development mode. This is now fixed.
-
Fixed: Folder filtering was not being applied in real time development mode. This is now fixed.
-
Fixed: Improved performance in real time development mode.
Released 2020.06.10
-
New: Added debug profiler support
-
New: We now return NLP debug data in RTT mode
-
Fixed: There could be a crash on RTT subset search with no results. This is now fixed.
Released 2020.05.27
-
New: Added support for RTT variables and actions
-
New: Added support for RTT dynamic settings info
-
New: Added support for RTT dialogs
-
Fixed: There could be a crash when tracking uncompiled RTT content. This is now fixed.
Released 2020.04.29
-
Fixed: We fixed the formatting for multiple attribute groups.
-
Fixed: We eliminated double-logging of user questions within subsets.
Released 2020.04.15
-
New: Added minScore, cutNext and cutLast to general settings
-
New: Support for basic RTT features
-
New: Support for RTT forms
-
Fixed: Legacy ticket creation tracking feature
Released 2020.04.01
-
New: There are two new supported languages: Macedonian and Serbian.
-
Updated: We updated Inbenta rate-limits.
-
Fixed: The API now returns the list of unknown words detected in the user input.
Released 2020.03.18
-
Updated: Improved list type variables detection
-
New: There are four new supported languages: Bulgarian, Croatian, Icelandic and Lithuanian.
Released 2020.03.04
-
New: There are four new supported languages: Albanian, Bosnian, Latvian and Slovenian.
-
New: Accepting a time zone setting to return datetimes on a user's time zone in the conversation history.
-
Updated: form field names are now deprecated.
Released 2020.02.19
-
New: Added features and more flexibility in Chatbot actions to send and receive variables from webhooks or callbacks
-
New: Disable or customize webhook error messages
Released 2020.01.22
-
Updated: Improved efficiency when calculating the semantic weight of user inputs
Released 2020.01.08
-
New: Improved info returned in the
conversation/history
API endpoint. In user interactions with revisitable links, the Direct Call name is returned instead of the non-user-friendly revisitable link value. In addition, user interactions with Option Id or Direct Call name now also return the message attribute instead of an empty string. This message is automatically inferred from Option Id or Direct Call name of the content.
Released 2019.12.18
-
New: The engine now returns the folder list hierarchy in
POST conversation/variables/options
and conversation/message
when a folder list variable is asked to the user.
Released 2019.12.11
-
New: You can now enable or disable the "show end form" message with
show_end_form
.
-
New: You can now identify every intent and encrypt the intent identifier in revisitable links.
-
New: The
conversation/history
endpoint now returns sidewindow text.
-
Fixed: A bug was found where revisited contents with non existent dynamic redirect were not returned. This is now fixed.
Released 2019.11.27
-
Fixed: A bug was found with sending the conversation history to webhooks. This has been fixed and attaching the history to webhooks is now possible when an internal setting is enabled.
Released 2019.11.13
-
New: We reduced the webhook timeout period to 30 seconds. This is because the API already had its own gateway timeout of 30 seconds, and PHP was still trying to execute webhooks after this period expired, and consuming resources. This reduction does not affect the Chatbot end-user experience and helps reduce the strain on our own system.
-
New: We added four new tracking events for registering when a user starts or abandons an intent and when the Chatbot fails to respond or responds to an intent.
-
Fixed: A bug was found where the Chatbot was ignoring callbacks in the "No Results" content, even though callbacks were working in all other contents. This is now fixed and you can execute callbacks in the "No Results" content as intended.
Released 2019.10.30
-
Fixed: A bug was found where the API was not executing Direct Calls when the Direct Call name was made up entirely of numbers. As a result, Sessions > Details and Chatbot Debug were registering a Failed Direct Call. This is now fixed so that even if your Direct Call name is "3" the API will correctly execute it.
Released 2019.10.02
-
New: Added trusted webhook flag to actions.
-
New: Tracking of JS Callback.
-
Fixed: Validate option ID in chained and multiple behaviors.
Released 2019.09.18
-
Added: New language detector endpoint.
Released 2019.09.18
-
Fixed: We updated the way the Logs Obfuscator works to make sure that even very open regular expressions have no adverse effect on the Chatbot Session Details screen.
Released 2019.09.04
Released 2019.08.21
-
Fixed: A bug was found where, when the dynamic setting TRACKING had an object in a content with an action, the Engine tracked the custom event twice: before showing the action and after. This is now fixed so that the event is only tracked once.
Released 2019.08.07
-
New: When the "TRACKING" dynamic setting is set in a content, a custom tracking is performed in the Session details.
-
New: Exposing Date header in CORS
-
New: Improved report of failed webhooks (added status code and new error types)
-
New: "No results" tracking
-
New: Handler in order to invalidate webhook responses setting invalid chatbot message responses
Released 2019.07.24
-
New: The new “Dynamic Redirect” feature allows you to configure the Chatbot to proactively show a particular dialog when it returns another intent as a matching result. This is particularly useful in dialogs, where the botmaster can proactively launch a hotel booking dialog, for example, after the matching process returns any content related to hotels or travel packages. This function allows you to reuse one or several dialogs across intents.
-
New: You can now set Chatbot to return 1-by-1 options without introduction text.
-
New: Added tracking of the number of options in multiple and chained results.
Released 2019.07.10
-
New: There is a new endpoint in
DELETE /conversation/variables
. This endpoint allows you to reset variable values to default value.
-
Fixed: A bug was found where sometimes, after a search with multiple results, the engine tracked a subset search when an additional input matched to external results. This is now fixed.
Released 2019.06.26
-
New: Action field data is now also returned in basic data fields (Previously, it could only be returned with advanced fields such as the date picker).
-
Fixed: Input fields are now disabled when the bot asks if the user wants to abort the form. They were previously disabled after the user confirmed the intention to abort.
Released 2019.06.12
-
New: Chatbot can now track intent data
-
New: Added
externalId
and title
attributes to external contents tracking.
Released 2019.05.29
-
New: Answers now return intent score info on intent start
-
Fixed: A bug was found where chained transitions were tracked with title as transition text. This is now fixed and they are tracked correctly with transition text.
-
Fixed: A bug was found where check questions with multiple values were not correctly tracked. This is now fixed.
-
Fixed: A bug was found where the engine was tracking bot message on actions without fields. This is now fixed so that it does not track bot messages, since no message is returned in these cases.
Released 2019.05.15
-
New: The Chatbot API can now track information about the SDK version used as part of the User information.
-
New: Ratings are now added to the Session tracking and visible in the "Session Details" section of Dashboards.
-
Fixed: A bug was found where, when a user rated a dialog node negatively, the click to the node was logged incorrectly in the new "Sessions > Details" and the "Ratings" dashboards. This is now fixed: the child node click is logged as "User click" and the rating is logged with the source "Chained Contents", without a user question.
Released 2019.05.07
-
New: Added support for Chatbot General Settings > Webhooks global configuration.
-
New: Added support for making dialog button links revisitable when contents contain a direct call and the links appear before automatic transitions or in dialogs without automatic transitions.
-
Fixed: A bug was found where, when a content returned had a webhook or JS Callback and no data fields, the API did not return the answer text. However, the new Chatbot Sessions > Details screen was registering these answer texts. This is now fixed and answer texts that the bot does not return are no longer shown in the Sessions > Details Event Data column.
-
Updated tracking information in the API Routes
Released 2019.04.17
-
New: Tracking improvements effort for the new "Session details" screen in the Chatbot App.
-
New: New endpoint for the Chatbot API that allows you to capture multiple variables. For example, when a user is looking to purchase sneakers, you could capture the variables "color = red" and "brand = adidas" in a single request. For more information, see the new endpoint
POST /conversation/variables/multiple
in the API Routes documentation.
-
Fixed: Tracking codes and no-rating flag are now properly returned on follow-up direct answer mapped as a content
-
Fixed: Returning Variable not found error message when trying to set a variable using API endpoint
POST /conversation/variable
-
Fixed: The No subset match direct answer was tracked twice. This is now fixed and it is only tracked once.
-
Fixed: Reordered variable tracking when clicking multiple options button
Released 2019.04.03
-
New: New
sys_total_negative_ratings
system variable to count negative ratings in a conversation.
-
New: The
/conversation
endpoint of the Chatbot API now returns not only the sessionToken, but also the sessionId. This allows developers to retrieve and use sessionId to filter results in Dashboards.
-
Fixed:
sys_total_negative_ratings
system variable did not work as expected of the rating code was sent as a string. This is now fixed.
-
Fixed: Variable default capture on actions was previously tracked twice. This is now fixed to only be tracked once.
Released 2019.03.20
-
New: Dynamic List Variables. Bot masters can now create list variables with potential values provided during the conversation by webhooks. For example: after receiving the end-user's user login info the bot can launch a webhook to retrieve that user's shipping record. The bot can then offer the user a selector or a series of buttons prompting the user to choose one of these records. For full documentation, see Using a webhook to capture Multiple-value variables in Webhooks and the articles Using Multiple-value variables and Using Dynamic List Variables in the Inbenta Help Center.
-
Fixed: The "no-question" direct answer was always tracked when a conversation started with an empty message. This is now fixed.
-
Fixed: In some situation, errors in the capture of variables by webhooks were not tracked correctly. This is now fixed.
-
Fixed: A bug was found where the "follow-up-question" System Message, which the Chatbot shows after the user repeatedly enters incorrect information in a single form field, was being logged twice in Sessions > Details. This is now fixed so that it is only logged once.
Released 2019.03.06
-
A bug was found where variables that were set with the detection scope "Validation only" with default values were not actually set with these default values at the start of each conversation. This is now fixed.
-
When using an automatic transition in the Dialog Manager, we usually add at least one "condition". This makes sure that the Chatbot transitions to the following child node based on the value of a variable. We must always include a "default" node in case none of the previous conditions is met.
Our internal error reporting system showed that in a single project, the Chatbot returned an "Oooops" message and was then unable to continue the conversation. This was caused by an automatic transition that had only the "default" condition configured. We also found two other factors: the root content was the Welcome content, and it did not have an Action (form).
Since the objective of this configuration is to respond with two answers (the root node's answer and the child node's answer), a workaround was found where we added a second Answer Text object to the root content and removed the automatic transition followed by the default node. This issue is now fixed and contents can be followed by the default node only, however if the desired outcome is to achieve two responses, we recommend using multiple Answer Text objects, as the performance will be slightly faster.
Released 2019.02.20
-
Fixed: Ratings for chained contents, related contents and revisitable contents
-
New: Error tracking
Released 2019.02.06
-
New: Webhook tracking
-
New: Action fields tracking
-
New: NLS subset tracking
-
Improved: Tracking of revisitable related contents
-
Fixed: Invalid click and rate tracking codes now throw a 422 HTTP status code instead of a 500 code (For more information, see API error codes).
-
Fixed: Issues were found when capturing user input with umlauts using regex. This is now fixed.
-
Fixed: The Welcome message was being logged twice. This is no longer the case and it is now only logged once as expected.
-
Fixed: Setting the Chatbot with nonexistent profiles and using Direct Calls no longer causes an error.
Released 2019.01.23
-
New: Added variable-based content folder filtering.
-
Fixed: Improved missing dbms failover.
Released 2019.01.09
-
Clicks on related contents are logged now with the data key CLICK_RELATED.
-
The "Fromsite" filter in the Dashboard tabs is renamed to "Environment", and you can now filter Dashboards by Development and Preproduction environments in addition to the Production environment.
Released 2018.12.19
-
Fixed: There was a conflict with legacy form databases that sometimes caused content to be evaluated as an action with no data fields and not as a simple content. This is now resolved.
-
Documentation: We have updated the
actionFields
section of the Chatbot API Answer types to include information about the "datePicker" fieldType. The API returns this information when a date variable is used in Actions.
Released 2018.12.12
-
Setting the SDK to use the Development environment no longer causes a crash after the user clicks on a related content.
-
Added a new function to the Chatbot API that facilitates changing the conversation token signing keys.
-
Changed conversation session signing key to a new one.
-
Fixed: Setting the Chatbot with nonexistent profiles and using Direct Calls no longer causes an error.
Released 2018.11.07
-
Fixed: The data key "START" was logged as if it were in a production environment even though the Chatbot is in the development environment. Now it will be logged as "external test".
-
POST conversation/variables also accepting numeric values.
-
In order to track custom user_info and USER_BROWSER and not send extra API requests, we have added a "tracking" parameter in the conversation endpoint payload, with "userInfo" attribute in it. The keys defined inside this attribute will be tracked as USER_INFO.
-
PHP session storage will now expire after 30 minutes of inactivity instead of 24 in order to standardize this session expiration time across all products.
-
Hotfix: Updated one of the libraries used by the API to ensure smooth transitions to/from Daylight Savings Time.
Released 2018.10.17
-
Added: New
POST conversation/variables/options
endpoint will return variables with a set of predefined options.
Released 2018.10.03
Released 2018.09.19
-
Added: Support for System Variables
-
Added: Support for Non-Matchable contents
-
Added: Support for No-results Dialog
Released 2018.09.05
-
Added: Some Direct Answers can now be replaced by contents with the correctly configured Direct Call name. For more information, see Customizing direct answers.
-
Added: Automatic transitions now support new logical operators ("<", ">", "<=", ">=").
-
Fixed: Validation exceptions 422 were returning an HTTP status code 500; now they return a 422 HTTP status code.
-
Fixed: Direct Calls with empty message were failing. This was fixed.
-
Fixed: Automatic transitions were not working after a button transition. Now, they are allowed.
-
Fixed: Automatic transitions were not working after another automatic transition. Now, they are allowed.
Released 2018.08.08
-
Added:
userInfo
endpoint to API.
Released 2018.07.25
-
Added: Direct Calls now allow bot masters to force load a specific content in Chatbot.
-
Fixed: Ratings are only shown after the associated form action in instances with the Ratings dynamic setting created.
-
Improved tracking response to invalid tracking code: it now returns a 422 error instead of a 500, and provides more feedback with a dev-friendly message
Released 2018.07.11
-
Added: Variable captures in different scopes: always, within intent, form fields only and validation only
-
Added: Variable captures using regex
-
Added: Variable capture using lists
-
Added: Variable capture filtering using string, number or date data types
-
Added: Different behaviors for form fields: always ask, skip, confirm
-
Added: Use captured variable values when displaying contents
-
Added: Obfuscate in UI sensitive type variables
-
Added: Ratings display configuration using RATINGS dynamic setting: true, false, auto. (See Content Ratings for more info.)
-
Added: In multiple results from a dialog, do not show children if there is a parent with higher score
-
Added: New logic that permits the use of variable values within content attributes
-
Added: Webhooks send conversation session id
-
Added: PolarQuestions and MultipleChoiceQuestions returning attributes
-
Added: Get variables endpoint
-
Fixed: Forms return rating on end form instead of between answer and first form question
-
Fixed: no-subset-match direct answer in Catalan
Released 2018.05.30
-
Fixed: log ratings properly when multiple contents are shown
Released 2018.05.16
-
Fixed: API not returning rate code (hotfix)
-
Fixed: use default timezone in logs
-
Fixed: corrections in exit labels in all languages
-
Added support for grouped attributes
Released 2018.05.02
Released 2018.04.18
-
Added: support for multiple external results (KM and Search API)
-
Added: new attribute source to answer output, with type (chatbot/external_km/external_search) and name (base name defined in Federated Bot configurator)
-
Fixed: intents with no results are logged
Released 2018.04.04
-
Federated Search
-
Updated to AIML v3. It compares the scores from KB and AIML in order to return the best matching.
-
Added: After a subset of options with no matchings, bot will search in all bases than just Chatbot contents
-
Added: New direct answer on skipping all check questions (all-options-discarded)
-
Added: Generic bot messages now return specific flags in answers
-
Fixed: Added translations for 'yes' and 'no' options of check questions
-
Fixed: problems with AIML logs
-
Fixed: Data2Extract variables with non-lowercase names can detect pre-filled variable values