Skylar AI 2.4.0 Release Notes

Skylar AI is a software services suite powered by artificial intelligence (AI) that automatically manages, anticipates, and prevents IT incidents. Version 2.4.0 of Skylar AI includes updates to Skylar Advisor and Skylar Analytics. Please note that Skylar Advisor currently requires Skylar Analytics.

The following table lists the minimum and recommended versions of ScienceLogic software required for Skylar AI and its components:

Product Minimum Version Recommended Version
Skylar One 12.3.2 12.5.20 or later
AP2 8.18.43-81 (Jelly Bean) 9.13.1-14 (Quesito) or later

 

If you are upgrading from Skylar AI version 2.2.1 to version 2.4.0, ScienceLogic strongly recommends that you upgrade from version 2.2.1 to 2.3.1 first, and then upgrade to 2.4.0. You will also need to run a reset script before deploying 2.4.0 if you received events for your Skylar Advisor instance. For more information, see Upgrading from Skylar AI Version 2.2.1.

What's New in Skylar AI 2.4.0

This release includes the following updates, which are covered in more detail in the sections below:

  • The Preview Features page in Skylar Advisor, which gives you the option to enable one or more new features. These features are stable, but ScienceLogic is actively developing and improving them in dialogue with users.
  • Skylar Advisor can create an Advisory and send data about that Advisory to Skylar One as an alert
  • Skylar Advisor can also create ServiceNow records from Skylar AI Advisories
  • High Availability configurations were added to this release to ensure Skylar AI components remain up and running
  • Additional features and addressed issues for both Skylar Advisor and Skylar Analytics

Features in Version 2.4.0

This section lists the new features that are included in this release of Skylar AI.

Skylar Advisor

  • When Skylar Advisor creates an Advisory, it also sends data about that Advisory to Skylar One as an alert. If you have an event policy set up to handle these kinds of alerts, Skylar One will generate an event that includes the Advisory title, summary, and a direct link to the Advisory in Skylar Advisor for streamlined investigation.

  • Enhanced the logic for creating Advisories, allowing for more precise event matching and severity assignment.

  • Skylar Advisor supports additional third-party knowledge imports on the Corpus page, including Confluence documents and generic JSON documents.

  • Added a "[Generated by Skylar AI]" tag at the bottom of articles created with the Knowledge Base Builder, integration messages sent to Slack, Microsoft Teams, and ServiceNow, and in the footer of Skylar Advisor as needed.

  • The following elements were added to the Skylar Advisor user interface:

    • The Skylar AI version number now displays on the login page.

    • The Dashboards page includes a new filtering button at the top of the page that lets you view all activities, only your activities (My Work), or archived activities.

    • The time range selector near the top of the Dashboards page lets you move week-by-week through your past activities. You can click the selector to update the time range, from six hours to 14 days, or you can select a custom range.

    • The Dashboards page provides access to Predictions, which are potential events that might occur in the near future if no action is taken. These predictions are generated using data from Skylar One and Skylar Analytics. Currently, Skylar Advisor generates predictions only from file system data.

    • The Dashboards page now includes a Services tab to display event data for Skylar One business services.

    • The Data Lens page has a new Signals tab, which replaces the Event Groups tab from previous release. The Advisor tab on the Data Lens page was also removed. Signals group correlates events from your monitored environment for triage, and can be rolled up into an Investigation. Dashboard heatmaps anchor those investigations visually; here, they're ordered by time.

    • The Team Analytics page from previous versions was renamed Customer Insights. The Customer Insights page ( > Advisor > Customer Insights) is an interactive, read-only page that displays data about the activities of all users on that system and instance.

  • The Preview Features page ( > Account Settings > Preview Features) contains a set of new features that users can toggle on or off so they can evaluate those features. These features are stable, but ScienceLogic is actively developing and improving them in dialogue with users. These toggles apply only to the current instance, not the entire Skylar Advisor account. 

    Click the "Play" icon next to each preview feature to watch a short demo of that feature. Also, some features might require you to refresh your browser before they appear.

    The Preview Features page includes the following options:

    • Selecting the Agentic option on this page enables the Agentic Control Plane page ( > Account Access > Agentic Control Plane). This page lets you monitor the AI agents in Skylar Advisor, enabling robust data-lake querying and controls to design agentic flows and govern the tools your agents are allowed to use. You can view a list of agents and turn them off if needed, and you can also monitor how many tokens an agent is using over time.

      To control and enforce the behavior of the Skylar Advisor AI agents, ScienceLogic has placed two gates in front of every agentic capability. The first gate is based on the release, where ScienceLogic only releases capabilities that have been vetted as production-ready, and anything unpublished is invisible to the agents. The second gate is based on customer opt-in, where a Skylar Advisor administrator must deliberately enable an agent before users can access it. As a result, a Skylar Advisor AI agent does not run unless it has cleared both the ScienceLogic release gate and the explicit opt-in by the site administrator.

    • In the Integrations section, selecting one or more of the options enables the Integrations page ( > Account Settings > Account Access > Integrations). This page lets you create integrations with Slack and Microsoft Teams using webhooks. These integrations let users share a Skylar AI conversation with other users by using the Share a Conversation icon () and the Collaborate panel.

      You can also enable ServiceNow Outbound to create ServiceNow records from Skylar AI Advisories.

    • In the Conversations section, you can enable one or more of the following features related to the Ask Skylar feature:

      • Ad-hoc File Upload. Lets you drag and drop one or more text files onto the Ask Skylar text box so you can ask Skylar questions using those files as context.
      • Trend Analysis. Adds a new panel to the Dashboards page that charts conversation volume, outcomes, and trends over time for a quick summary of your Skylar Advisor activities.
      • Conversation Audit. Lets you view the full token spend and timing of an entire conversation thread for cost and performance review.

    • In the Work Items section, you can enable the Claim & Close option to add the ability to claim an Advisory or an Investigation (assign it to yourself) and close an Advisory or an Investigation when you have addressed it. This option also adds additional filters to the Dashboards page: Open, Claimed, and Closed.

    • In the Platform section, you can enable the Active Session Tracker option, which adds a new button in the top navigation bar of the user interface, for an Owner user or higher. The label for this button lists the number of active user sessions along with the total number of allotted user sessions. Clicking the button displays a list of active sessions, active users, and a Clear Active Sessions button to clear all active sessions.

Skylar Analytics

  • Added a default time range of seven days to charts that were missing this value on the timestamp columns (FSDataTime, IFDataTime, BsvcDataTime). This default setting helps reduce unnecessary database load.

  • Optimized performance by replacing manual cleanup for the Advisory_alert_queue table with a Time-to-Live (TTL) policy. This change prevents backlogs that previously led to service degradation.

Skylar AI Platform

  • For SaaS deployments, enabled High Availability (HA) by default for the Skylar AI Tenant service, the OpenTelemetry (OTel) Collector, Postgres, and ClickHouse Keeper. For more information about adjusting or turning off HA settings, see Configuring Skylar AI System Settings.

  • This release includes a number of updates that help to reduce overall execution time and improve performance of all Skylar AI components.

Issues Addressed in Version 2.4.0

This section lists the issues that were addressed in this release.

Skylar Advisor

  • Addressed an issue with the Instances filter where the Dashboards tab and the Data Lens tab displayed identical issues, devices, and organizations across all connected instances. The Instances filter now correctly scopes data to the selected instance. (Jira ID: SKY-4175)

  • Addressed a filtering issue in the Services Heatmap view where device filters failed to update the result set. Additionally, resolved a double-counting issue where events spanning multiple business services were incorrectly aggregated. (Jira ID: SKY-4265)

  • Resolved an issue where the Customer Insights dashboard failed to report Advisory counts for specific instances, while that activity was displayed in other views. (Jira ID: SKY-4273)

  • Addressed an issue where Advisory titles did not match the content of the associated reports. (Jira ID: SKY-4301)

  • Adjusted the scroll bar in the Skylar Advisor user interface to keep it from flashing white in dark mode. (Jira ID: SKY-4229)

Skylar Analytics: Data Visualization

  • Resolved an issue where the sync-all-datasets.sh script exited successfully even when some tenant datasets failed to sync, causing upgrades to be incorrectly marked as successful. The upgrade process now accurately reflects failures during dataset synchronization. (Jira ID: SKY-3756)

Skylar Settings

  • Improved the password set and reset process in the Skylar AI user interface to ensureusers receive accurate context when a link is invalidated or when active session limits are reached. These updates replace the generic "invalid link" messages for these situations. (Jira ID: SKY-4164)

To view the Skylar AI documentation online, see the Skylar AI Product Documentation.

Upgrading from Skylar AI Version 2.2.1

If you are upgrading from Skylar AI version 2.2.1 to version 2.4.0, ScienceLogic strongly recommends that you upgrade from version 2.2.1 to 2.3.1 first. After checking to make sure the 2.3.1 release is stable, upgrade to 2.4.0.

If you are currently running Skylar AI version 2.2.1 and you plan to upgrade to this release, you will need to run a reset script for Skylar AI before you deploy version 2.4.0 or later. This script ensures that new events are given proper consideration when forming new Advisories in Skylar Advisor, and the script avoids weighting these events negatively based on prior processing.

You only need to run this script if you are using Skylar Advisor version 2.2.1 and you received events for your Skylar Advisor instance. If your instance had no event traffic on version 2.2.1 or if you are a new customer, you can skip this section. You can also skip this section if you are only using Skylar Analytics, without Skylar Advisor.

Running the Reset Script

Before you deploy Skylar AI version 2.3.0 or later, you or your ScienceLogic contact will need to run the skylar_ai/event_cluster/run_event_table_reset_for_ver2.3.sh script once to reset the relevant tables.

To access and run the reset script:

  1. Before deploying version 2.4.0 or later, add the following values to the global secton in values.yaml:

    skylar-advisor-pipeline-event-cluster:
  enabled: false

skylar-advisor-pipeline-event-adv-gen:
  enabled: false

skylar-advisor-pipeline-event-rule-scan:
  enabled: false

skylar-advisor-pipeline-event-temporal-analysis:
  enabled: false

  2. Deploy the rest of the services. The four services above will stop running.

  3. SSH to any other service that has the skylar-ai repository mounted with python, such as the skylar-advisor-api pod. Any of the api services will work if there are more than one.

  4. Change directories to "/workspace/skylar_ai/skylar_ai/event_cluster/run_event_table_reset_for_ver2.3.sh.

  5. Run the reset script once with your tenant ClickHouse instance name:

    ./run_event_table_reset_for_ver2.3.sh <clickhouse_instance>

    For example:

    ./run_event_table_reset_for_ver2.3.sh it_it_default

  6. Review your output. You should see lines like the following in your output:

    DROP TABLE IF EXISTS it_it_default_ai3.AdvisoryOut
DROP TABLE IF EXISTS it_it_default_ai3.AdvisoryOut_staging
...
...

  7. Return to values.yaml and remove the code block you added in step 1.

  8. Redeploy all the services again.

Additional Updates from Skylar AI 2.2.1

The following changes from version 2.2.1 were made in Skylar AI 2.3.0 and later:

  • Centralized the Skylar AI ClickHouse configuration into skylar-clickhouse-chart and upgraded the ClickHouse Helm chart to version 9.1.1. This update includes the deployment of an audit logger with each ClickHouse instance and the addition of Time-to-Live (TTL) settings to all system tables.

  • Also, the shape of the yaml block for the ClickHouse configuration was updated:

    • Skylar AI 2.2.1:

      clickhouse:
 ....

clickhouse-advisor:
 ....

    • Skylar AI 2.3 and later:

      skylar-clickhouse-chart:
  clickhouse:
 ....

clickhouse-advisor:
  clickhouse:
 ....

Known Issues

This section lists the known issues for this release.

  • The Agentic Control Plane page is temporarily retrieving sample data and does not reflect the actual device data. (Jira ID: SKY-4179)

  • Dynamic Application objects with special characters in the name might conflict with other objects in Skylar AI reporting tables. The workaround for this issue is to modify the Dynamic Application objects so that the name does not contain a special character. (Case: 00617628) (Jira ID: SKY-3305)

  • A Skylar AI instance can ingest telemetry from multiple external sources, but only oneSkylar One system should be configured to send telemetry to that Skylar AI instance to prevent data collisions and potential data corruption. To avoid this situation, do not reuse a Skylar AI authentication token with more than one Skylar One systems that are sending data to that Skylar AI instance. (Jira ID: SKY-3538)

  • If the session inactivity timeout value for a Super User or a standard user is adjusted on the Authentication page ( > Account Access > Authentication) of Skylar Settings, the affected user does not get logged out after the new timeout value expires (if that user is currently logged in and remains inactive). (Jira IDs: SKY-3734, SKY-3732)

  • If you had customized tables and views in an older version of the Data Visualization component of Skylar Analytics, that data will not display correctly if you upgrade to Skylar AI version 2.1.0. Any tables and views created in the local database require that the "InstID" and "OrgID" columns contain data for related datasets, charts, and dashboards to be functional. You should not experience this issue if your tables and views in the local database contain values in the "InstID" and "OrgID" columns. (Jira ID: SKY-3824)

  • If an Administrator user removes the "Event Trends (Sample)" dashboard on the Manage Bundles tab of the Dashboards page, the dashboard still appears in the Data Visualization component of Skylar Analytics. Also, if the "Interface Explorer (Sample)" dashboard is added using the Dashboards page, that dashboard does not display in Data Visualization. (Jira IDs: SKY-3739, SKY-3737)

  • Skylar AI uses unique PowerPack naming conventions for creating datasets in the Data Visualization component. If you change the name of a PowerPack, Skylar AI triggers a new dataset, which will cause charts to fail and lose the historical data for that PowerPack. At this time, ScienceLogic recommends that you do not rename any PowerPacks in Skylar One if that Skylar One system is connected to a Skylar AI environment.

  • When Anomaly Detection trains on a metric whose declared unit is "%", the upper bound of the AD confidence band is constrained to 100 based on the assumption that percentage-unit metrics fall within 0–100. However, this setting does not account for metrics that could return values above 100%. For more information, see the following Knowledge Base Article (KBA) at the ScienceLogic Support Center: https://support.sciencelogic.com/s/article/21269. (Case: 00643786) (Jira ID: SKY-3763)