V‑Spark Version 3.4.3 provides improvements in a number of areas including analysis metrics, API usage, metadata handling, search, and general system operation.

Daily statistics are now generated in scheduled batches. This replaces the old behavior of generating statistics one file at a time.

The database queries that are used during the calculation of Application summaries have been optimized to be more computationally efficient.

1. Improved Diarization Scores - Diarization now uses a new scoring algorithm that better represents how well mono audio was split into two speaker channels. Scoring used to be in relation to previously uploaded transcripts, but is now independent of transcript history.

2. Better File Revision Tracking - The system now allows more than one transcript to have the same filename and received date. Duplicated transcripts used to be given a different timestamp to differentiate them from previously uploaded transcripts. Duplicates are now differentiated by transcript ID (tID).

3. Improved Loading of System Configuration - System configuration is now stored in a directory of multiple files instead of a single file. Configuration files in the vspark.config.d directory are read in alphabetical order at server startup. Legacy configuration files are still read, but the files in vspark.config.d always take precedence.

4. "Last Generated" Time Display - Statistics are now shown with the time they were last generated. The display may lag behind actual statistics, as the analysis only runs every five minutes, (This interval is configurable.)

5. New /sysinfo API - The REST API now has a /sysinfo call that returns system information.

6. Improved /config/users API - New user accounts created using the /config/users API can now be configured to use external authorization methods. Note that existing users cannot be reconfigured in this way.

7. Additional Fields Supported in Search API - The Search API feature allows four additional output fields: "tags", "folder", "url", and "es_doc_id" (Elasticsearch document ID).

8. Search Results Now Display Custom Metadata - When doing a search for custom metadata, the values of the custom metadata field being searched are displayed in the search results. It is no longer necessary to click through to the file details to see how results match the custom search criteria.

9. Custom Metadata Now Appear in Filter Dropdown - If your Application is linked to a folder with custom metadata fields, those fields now appear directly in the "Filter" drop-down menu instead of a separate "Custom Metadata" menu. This makes metadata fields easier to find and faster to utilize.

10. Unicode Support in Metadata - When uploading audio files with metadata, fields in the metadata may now contain any valid Unicode values, except for reserved characters Colon (":", Unicode 003A) and Equals Sign ("=", Unicode 003D).

11. Improved Login with External Authorization Methods - Improved linking for systems that have external authorization enabled. Contact support@vocitec.com if you are interested in this capability.

12. Logging of Unsuccessful Login Attempts - For security purposes, the activity log now tracks all unsuccessful login attempts along with the reason the attempt failed, such as incorrect password or username, unverified user account, user account disabled, or unmapped user account.

13. Improved View of User Permissions - In the Company table of the Accounts Settings page, there is now a "Permissions" column that contains a "View users" link. Follow this link to view a list of all users who have permissions that apply to this company and its organizations. Permissions are displayed in a dialog that shows all V‑Spark users with permissions, whether those permissions are explicit to the company you chose, or implicit in the permissions those users have in the system. If a user's permissions in one company or organization are based on the permissions they hold in another company, the name of that other company is displayed in parentheses next to their username. By default, permissions for the entire company are displayed. To show permissions for a specific organization within the company, choose the name of the organization from the drop-down menu above the list.

14. Notification of Usage Limits - The system generates a system notification when a company nears their usage limit, and another if the usage limit is reached. The warning level is configurable, and defaults to 80%. These notifications are visible to system administrators and to users with write access to the affected company. Notifications are dismissible within V‑Spark, and will only display once per day. These notifications are also sent via email to system administrators and company administrators, and recorded in the activity log.

15. Folder processing can now be paused via the API - Using the /config/folders API, you can now pause the processing of a Folder by POSTing a JSON configuration file for the Folder that has the mode property of the Folder set to the value paused . Resume processing by POSTing JSON for the Folder that has the mode property set to active .

This section identifies specific issues that have been resolved in the 3.4.3 release of V‑Spark:

1. Deletion Jobs Now Span System Restarts - If the system is shut down or restarted while a deletion job was executing, the job was not being completed. Incomplete deletion jobs, whether user-initiated or automatic, are now re-queued for execution when the system starts.

2. Improved Handling of Application Scoring Failures - Previously, uploaded files that failed scoring would be moved to the error directory, even if other analysis succeeded. These files will no longer be moved to the error directory. Scoring failures are now recorded in the activity log and backend logs, showing which file failed scoring, and that none of the Applications that use that file will have scores for that file.

3. Transcript Processing is Now Atomic - If an Elasticsearch or MySQL failure is encountered during transcript loading, all related updates are rolled back cleanly to preserve data integrity.

4. System Mode Changes are Now Logged - System Mode changes between "normal", "safe", and "readonly" were not being logged, which made diagnosing trouble more difficult. The system now logs when mode changes occur, and (depending on warning level) sends email to system administrators.

5. Trending Phrases No Longer Reported - Trending phrases were being included in daily and monthly reports, even if Trending was disabled. This has been corrected.

6. Appstats Now Returns Categories with Scores of Zero - The /appstats API was not returning any categories if the top category score was zero, even if the depth level was defined to return categories with scores of zero. This defect was corrected, and categories are now returned properly.

7. The /transcribe API was not processing audio files if they were uploaded from Amazon's Simple Storage Service (S3). The system was returning an empty reply. This defect has been corrected, and the system now returns a transcript as expected.

8. Files with Empty Utterances Now Update Properly - Uploaded files that result in an empty Utterances field were ignored during application reprocessing. This issue has been resolved and these files are now updated correctly.

9. Config Now Returns Correct Template Value - The /config/CO_SHORT/ORG_SHORT/apps/APPNAME API was not returning the correct template value for Applications that were created from templates. It was returning the "template" value of "custom" even for Applications that were created from templates. This issue has been resolved, and the system now returns the correct template value.

This section identifies fixes that were introduced in patch releases and hot fixes since the initial 3.4.3 release, and are therefore now included in the latest version of the 3.4.3 release. Maintenance versions and associated fixes are listed most recent to earliest:

1. Improved response handling for systems using V‑Cloud. Eliminated extraneous 503 errors generated when communicating with V‑Cloud to minimize unnecessary noise in Job Manager log files. (3.4.3-15)

2. Fixed an issue that would cause Chrome to auto-fill the search box on a folder page with the current username. (3.4.3-14)

3. Fixed a typo in the How it Works page image. (3.4.3-14)

4. Google Chrome 83 Support Update - Chrome 83 now blocks all downloads initiated from or instantiated in a sandboxed iframe by default. This update enables V‑Spark to override that default behavior and allow file downloads. (3.4.3-13)

5. Improved handling of email communication - Email notifications were updated to list email recipients in the BCC field of the message. Also, the TO field is now set to the application's global address. (3.4.3-12)

6. Improved support to AWS S3 downloads - Improved download retry logic to account for possible issues during network transport. (3.4.3-11)

7. Run V‑Spark on a Single Core - Enabled V‑Spark to run on a machine with a single core on its single CPU. (3.4.3-10)

8. Eliminated Temporary File Leaks - Improved cleanup of temporary files. (3.4.3-10)

9. Fix Application Reprocessing Failure - Improved mysql retry logic to reduce application processing errors. (3.4.3-10)

10. Audio Duration Update - Improved how a file's audio duration is determined. (3.4.3-10)

11. S3 Update - V‑Spark now fully supports signed S3 bucket requests for all AWS regions by adding support for the AWS Signature Version 4 Signing Process. Previously only Version 2 was supported. (3.4.3-10)

12. Partial Content Requests - Some browsers, such as Google Chrome, specify a desired content range for cases such as fetching a piece of an audio file to play/seek in its embedded audio/video player. V‑Spark now honors partial content requests on the part of a browser by responding with 206 HTTP code and the piece of audio that was requested. (3.4.3-10)

13. Elasticsearch connection management - Connections to Elasticsearch clients are now managed more effectively to prevent connection errors. (3.4.3-9)

14. Folder ASR options - Expanded support for folder ASR options to allow adding values with multiple elements via the /config/folders API and GUI. Previously, both methods only supported simple ASR options with string, boolean, or numeric values. Boolean ASR option values must now be formatted properly for them to be correctly recognized. The Boolean values true and false must be entered in all lower case with no quotation marks. Values containing capital letters or contained within quotes (such as True or "false" ) are interpreted as string values, and not as Booleans. (3.4.3-8)

15. Request API Content-type - The /request API status and details calls were returning output with the incorrect content-type value. The content-type has been restored to "application/json". (3.4.3-8)

16. Cloud Tokens Checked Properly on Company Update - Updating a company that uses cloud tokens was pausing all of the company's folders due to a token check that was not matching the tokens to the folder servers properly. Tokens are now checked properly, enabling companies to update without their folders being paused. (3.4.3-7)

17. Emotion Word Highlighting - Turning off "emotion" view on file details page will mitigate a known issue where phrase matches that overlap words marked with emotion are not highlighted. Now, opening a file details page from the application dashboard will turn off emotion by default, to improve the display of application phrase matches. (3.4.3-7)

V‑Spark 3.4.3 does not add any new announcements of upcoming changes and deprecation. Please see Announcements with 3.4.2 for announcements made with the 3.4.2 release.

1. Partial configuration updates for a company, organization, folder, application, or user using the REST API are not currently supported. Attempting to update a subset of a company or organization using the REST API currently returns HTTP error code 400. You must supply all fields for the configuration in order to update any field(s) using the REST API.

2. Application re-processing of more than 100,000 files in a single operation may time out.

3. Issues with Dashboard Displays

   • Files uploaded to a folder that is linked to a disabled application or that have not been re-processed will still appear in the Files View on the application dashboard

4. Issues with Search

   • Time-based search

     • Searches are turn-based and may return results for phrases that are outside of the requested time frame if those phrases are within the same speaker turn

     • Ad-hoc Search snippets in the Files View may display results outside the specified time interval

   • Exceptions to standard File Details highlighting:

     • Searches using (~n) may highlight phrases with the specified number of words in between each of the search terms, rather than between all of the search terms. For example, "get help phone ~4" may highlight phrases with 4 words between "get" and "help" and another 4 words between "help" and "phone". Searches using (~t) and (~n) will only be highlighted if the words matched appear in the order specified in the search tag.

     • Highlighting may be incomplete for matches marked with emotion

     • Highlighting may be incomplete for matches with intervening silence tags greater than 2 seconds

     • Matches for "@*" (such as when searching for email addresses) will only be highlighted until the first "." in the email address. Regular expression searches may be used to circumvent this issue.

     • Excluded application search phrases may be highlighted on the File Details page. Note that these exceptions only affect highlighting, and that all search results using the above characters will still be accurate.

5. Other

   • Users playing longer audio files via Firefox may encounter playback/syncing issues when jumping to different sections of the transcript. This is an external issue caused by the Firefox browser. Users who encounter this issue can simply wait for the browser to finish buffering the file before resuming their search.

   • App Reprocessing will encounter an error if the reprocessing occurs while an App or Category is being deleted. This error is recorded in the server logs.

   • Searches that contain both wildcards and spaces are not returning results.

   • Most punctuation characters are not supported in search strings.

   • The /search API allows searches for application categories that have hyphens and other restricted characters in their names, however these searches fail. Voci application templates that had hyphens and other restricted characters in their names have been modified to remove the restricted characters. All existing application category names should be modified to remove restricted characters.

   • If the server is restarted while files are waiting in the Callback queue, those files will not be delivered and the callback will not complete. Allow Callback delivery to complete before restarting the server.