24.01 Release notes

Instabase 24.01 is a major release that introduces new features, enhancements, and bug fixes.

Subsequent patch releases typically contain bug fixes along with testing, optimizations, and other minor internal changes.

Release 24.01.19

  • You can now include multiple apply classifier steps in flows. For example, you can use a first split classifier step to classify documents into general types, like identity documents and tax forms. Then, you can use a second split classifier step to subdivide these general types into more specific types, like passports and driver’s licenses.

Release 24.01.18

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 24.01.17

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 24.01.16

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 24.01.15

This version was not released.

Release 24.01.14

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 24.01.13

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 24.01.12

  • Source data is more accurately identified for extraction prompts due to improved OCR token matching.

  • When resuming a job from Flow Review, the status did not update if you stayed on the Flow Review screen.

Release 24.01.11

  • From the Flow Review dashboard, you can now bulk assign jobs to a reviewer.

Release 24.01.10

  • You can now visualize the progress of your flows on a file-by-file basis. Click into a file in your flow, and you can see any long-running steps that delay the completion of their flow.

Release 24.01.9

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 24.01.8

  • Annotations weren’t carried over when importing an annotation set from the file system into Solution Builder, or when exporting an ML Studio dataset in the same environment.

Release 24.01.7

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 24.01.6

  • In ML Studio, the field list failed to load in certain circumstances.

Release 24.01.5

This patch contains testing, optimizations, and other minor internal changes. User functionality is unchanged.

Release 24.01.4

This is the first generally available release of Instabase 24.01.

New features

Flow Review

  • Public Preview | Document-based review introduces a new method for managing human reviews where reviews are broken down into individual documents. This method provides more granularity in review assignment, and more flexibility for reviewers, because you can assign, review, or escalate individual documents, or open multiple documents from different flow jobs for review. Flows must be resumed as a separate step when all associated documents are marked reviewed.

    Document-based reviews are supported only for flows that are managed as a deployed solution. Document-based reviews can’t be exported, and they’re not included in metrics reported in the Solution Dashboard.

Enhancements

Deployment Manager

  • You can now use the Update base config image API to override the defined registry. The change persists until a new base config is applied for the specified service, at which point the image uses the registry defined in the Installer.

  • From the Infra Dashboard, you can now view the live Kubernetes configuration for a given Kubernetes element. See the element’s Config tab for the materialized configuration and the Live K8s Config tab for the live configuration.

  • Network policies have been added for the jobs underlying the Test Runner test post-install action (solution-diagnostic-artifacts) and the base model image upload post-install action (mktplaceupdate). Previously, if your security settings required network policies for all jobs, these actions wouldn’t run. mTLS traffic is not currently supported when running these jobs.

Observability

  • The legacy Prometheus service has been succeeded by the VictoriaMetrics vmagent service. The vmagent service is significantly better optimized, boasting up to a 7x improvement in memory optimization and approximately 2x improvement in CPU optimization, compared to Prometheus.

    • The vmagent service supports horizontal scaling, allowing you to expand its capacity by adding additional instances.

    • The vmagent service is deployed as a StatefulSet service.

    • Rollback procedures are included as a contingency measure in the event of any issues.

  • VictoriaMetrics vmalert service now handles the evaluation of alerting rules. This service operates as a sidecar alongside VictoriaMetrics, enhancing the alerting capabilities of the system.

Flow

  • Improvements to the Flow job service provide faster startup for running flow jobs, reduced memory consumption, and increased queue capacity. Specifically, internal testing demonstrated a 5x improvement in startup time and the ability to queue up to 18x more jobs.

  • The Python library pyxlsb is now available for use in UDFs. This library enables importing .xlsb files in a pre-flow UDF.

Flow Review

  • You can now optionally display provenance only for the selected field by enabling Only Show Selected Field Annotations in field list options.

Flow Dashboard

  • Public Preview | Flow logs now include a visualization of processing time for each file and step in the job, so you can quickly identify slowdowns. You can drill in to each file to further visualize time taken by each step for a specific file. To access flow execution visualizations, from Flow Dashboard, click Logs for any job, then select the Visualize tab.

  • You can now view and filter by priority in Flow Dashboard. A new Priority column can be enabled in Flow Dashboard settings, and in the filter sidebar, you can filter by a specific priority. Flow priority can be set when using the API to run a flow.

  • The tag search filter in Flow Dashboard, as well as the jobs/list API endpoint, is now case sensitive.

Bug fixes

Flow

  • License usage is more accurately counted for flows that include apply checkpoint steps. Previously, usage was counted only when a flow completed.

Flow Review

  • When you used the pop-out icon to open a document in a separate window, the first file in the file list was displayed, rather than the selected document.

Deprecations and removal

  • Flow v1 is deprecated and scheduled for removal in 24.10. Use Flow v3 instead.

  • T4 GPUs have been removed from the supported GPU list. The minimum supported GPU for machine training is now A10. T4 GPUs no longer meet minimum performance requirements for machine learning, so training performance is likely to be poor. If you are using T4 GPUs for training, transition to A10 GPUs.

  • Support for ABBYY is deprecated in 24.01. ABBYY will no longer be supported after July 2025. If you are using ABBYY OCR, switch to another OCR such as MSFT OCR, which offers significant performance and accuracy improvements over ABBYY.

  • Recipe Book projects are deprecated in release 24.01. Custom UIs created with .ibrecipebook files are still supported.

Deployment guide

  • Workload autoscaling was introduced as a public preview feature in release 23.07 and remains in public preview. As a public preview feature, workload autoscaling is disabled by default. Workload autoscaling has several infrastructure requirements. For instructions on enabling workload autoscaling during your upgrade, see the workload autoscaling feature documentation.