Updated the retry logic for handling database errors. V‑Spark now retries any database error with an error code between 2000 and 2099 (inclusive), along with the codes 9001, 9002, 9003, and 9006. Previously, V‑Spark retried only a more limited set of 2000-level errors.

Improved logging functionality. Log messages generated when detecting language model availability now include a URL for the request that generated the message.

Updated the sorting logic for columns in CSV exports generated from the Files View list. Columns in reports created with the Files View are now sorted consistently and independently of the order shown in the UI.

Columns are sorted alphabetically, from left to right, in two groups. The first group contains general columns like agent id, duration, and other audio record attributes. The second group contains application-related columns, the names of which begin with the app. prefix.

This change ensures that column order will be consistent for every report generated, including paginated reports.

Updated the Elasticsearch package included with V‑Spark to account for the security vulnerability associated with the Log4j security flaw. The voci-config-elasticsearch package included with V‑Spark 4.3.1, and initially released with version 4.3.0-2, allows users to upgrade to any minor version of Elasticsearch 7 with the caveat that versions newer than 7.16.2-1 have not been tested for use with V‑Spark.

The full list of service dependencies tested with various V‑Spark versions is available at the following link: Third-party service QA tests. The versions tested with this release are included with these release notes.

For more information about V‑Spark and the Log4j vulnerability, please refer to the following statement from the Voci support team:

1. Voci V‑Spark does not use Log4j directly, so we are not planning any updates to V‑Spark software for this issue.

2. V‑Spark does have an operational dependency on Elasticsearch. The Elasticsearch team has determined that there is no remote execution risk in version 7.x releases, and that a minor issue with potential DNS-based information exposure is easily mitigated with a basic configuration update. Out of an abundance of caution, Voci advises that the mitigation be performed.

3. The best solution to this issue is an upgrade to Elasticsearch version 7.16.2-1 or greater, which users can perform at any time convenient to them and their teams. Voci does not anticipate any issues with using Elasticsearch versions greater than 7.16.2-1 with V‑Spark, but those versions have not been tested by the V‑Spark Engineering team.

More information about the risks and remediation associated with Log4j and Elasticsearch can be found on Elasticsearch's website.

Increased the timeout threshold for jobs in the report worker queue. Jobs now wait in the report worker queue for a maximum of 1 hour before timing out and being requeued.

Added functionality to escape certain characters in the field names for exported CSV reports. Custom metadata fields in exported CSV reports whose names begin with = , + , - , or @ are prefixed with an apostrophe ' to prevent spreadsheet applications from interpreting these labels as formulas.

Improved error handling when Elasticsearch is unavailable. Transcript worker processes now pick up new jobs only when these processes detect that Elasticsearch is up and running. Previously, jobs accepted while Elasticsearch was unavailable were picked up and immediately placed in the errors folder.

Added example and reference files to V‑Spark's system configuration settings directory. The four new example files show system configurations for various use cases and are located in the config/vspark.config.d.examples directory. The reference copies of default configuration files are located in the config/vspark.config.d.examples/base.defaults directory.

Improved V‑Spark security to help protect against remote code execution and other vulnerabilities.

Added functionality to display and export columns for custom metadata fields and their values using the Show/Hide Columns dropdown in the Dashboard Files View. The new Custom metadata section of the dropdown list includes all metadata fields associated with the folder currently selected in the dashboard. If the dashboard is set to display all folders, the dropdown includes all custom metadata fields for every folder the user has permission to view.

Enabling custom metadata fields using the dropdown adds a column for each enabled field to the table of search results. If a record in the list of search results has a value for the enabled field, that value will display on the search result line in the cell associated with the custom metadata field column; otherwise, the cell is blank. Enabled fields and their values are also included in search result CSV exports.

As a part of this change, the Delete icon for search results displayed in the files view has been moved to the left side of each line in the results list to improve user experience. Previously, the icon appeared on the right side of each result line.

Improved user experience when tagging search results in bulk. The dashboard now displays a Saving... message while adding tags to multiple search results. Previously, when tagging a large number of files, the page would sometimes display a Successfully Added Tags message before the new tags were visible in search results.

Increased the timeout threshold for jobs in the dashboard worker queue. Jobs now wait in the dashboard worker queue for a maximum of 4 hours before timing out and being requeued.

Added log messaging for all requeued jobs. Requeued jobs generate a WARNING-level message with the date and time, type of worker process, and other details in backendWorker.log with the format used in the following example:

2021-12-01 18:01:00.108 TranscriptWorker 28387 WARNING  Job Requeued: Found inflight job (transcripts_f145c2a1-2a91-4cef-a245-ac4521e8c17b) with an expired TTL -- <requestid=06babe23-2ad5-44b2-8153-4ceed1f76cd8>

Changed the format for temporary Elasticsearch index names. Previously, V‑Spark assigned the prefix .vspark-temp- to temporary Elasticsearch indices created for application scoring. The default value for this prefix has been changed to @vstemp to account for changes to index name requirements in the next major version of Elasticsearch.

In addition to the prefix change, temporary index names now include date, time, and detailed request components. The new complete format for temporary index names is prefix_YYYYMMDD_HHMMSSz_requestid_orgshort_tid and includes the following components:

• prefix — The prefix configured in V‑Spark, which is @vstemp by default.

• YYYYMMDD — The year, month, and day on which the index was created.

• HHMMSSz — The hour, minute, second, and time zone for the index, expressed in UTC.

• requestid — The unique identifier for the request that generated the transcript.

• orgshort — The short name of the organization associated with the transcript.

• tid — The unique transcriptID identifier for the transcript.

Addressed an issue with startup verification of database names that contain both lowercase and uppercase characters. Previously, verification would sometimes fail when checking the names of database servers with lower_case_table_names set to 1, such as is common on Azure-hosted databases. For more information, refer to the external Azure database documentation.

The third-party Python and NodeJS packages, and the optional third-party Elasticsearch and database packages, used with V‑Spark are now offered as independent packages in the Voci repository.

• Python — voci-spark-pylib

• NodeJS — voci-spark-nodelib

• Elasticsearch — voci-config-elasticsearch

• Databases — voci-config-database

Improved V‑Spark security to help protect against configuration request forgery and remote command execution.