Skip to content

2.1.995

Latest
Latest

Choose a tag to compare

View all tags
@arthur-louisa arthur-louisa released this 11 Oct 18:18
· 23 commits to dev since this release
2.1.995
ed4eb2d
This commit was created on GitHub.com and signed with GitHub鈥檚 verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

馃殌 Arthur Engine Release

October 11, 2026

This release expands SIEM discovery capabilities with new Splunk Enterprise and Elastic Security connectors, introduces a standalone ML Engine discovery mode, and delivers a range of connector validation fixes and observability enhancements.

Discovery & SIEM Connectors

New Connectors

  • Added Splunk Enterprise as a SIEM discovery connector, enabling organizations to discover and monitor security events directly from Splunk, with robust TLS handling, job lifecycle management, and comprehensive error classification (#2404)
  • Added Elastic Security as a discovery connector that executes customer-authored ES|QL queries and publishes results as discovery records, with configurable TLS verification and automatic error classification (#2405)

Standalone Discovery

  • Introduced a standalone agent discovery runner mode for the ML Engine that operates independently of the Arthur Platform and GenAI Engine, scanning sources on a schedule and sending discovered agents to Splunk HEC or a webhook (#2427)
  • Bumped arthur-client to 1.4.2770 to enable Test Connection support in the ML Engine, including the required job kind, test-result models, and gcp_vertex vendor surface (#2415)

Connector Reliability & Validation

  • Consolidated URL validation across the Jamf, Splunk, and Elastic connectors into a shared address_problem function, ensuring malformed URLs are consistently rejected at setup time with clearer error messages (#2429)
  • The Jamf connector now validates base_url format before making requests, reporting malformed URLs as not_configured instead of generic provider errors (#2424)
  • Vertex AI discovery scans now correctly classify HTTP 401 and 403 failures as authentication and permission errors rather than generic provider errors (#2428)
  • Fixed Test Connection result classification so timeouts with no data correctly fail as TIMEOUT and configuration errors are no longer misclassified as vendor errors (#2417)
  • Manual discovery scans now tag their chained fetch operations with the scan's ID, allowing the Platform to correctly track scan completion before reporting results (#2425)

This release significantly broadens the Arthur Engine's SIEM integration footprint while making configuration and connectivity diagnostics more accurate and easier to troubleshoot across all discovery sources.

Observability & Evaluations

  • Trace annotations now correctly populate the evaluation type field for continuous LLM and ML evaluations, ensuring accurate metadata regardless of sort order or lookup method (#2431)
  • Added TypeSafe AI instrumentation support to the Python SDK via an optional typesafe installation extra, included in the all dependency group (#2423)
  • Fixed a bug where metrics and alert-check jobs could overwrite user-initiated check chains; the engine now stamps follow-up job IDs only onto the chain a completed job actually heads (#2407)

These improvements give users clearer visibility into which evaluations generated each annotation, expand supported instrumentation frameworks, and prevent scheduled runs from interfering with user-initiated checks.

Infrastructure & Governance

  • Added a CODEOWNERS human-review gate requiring approval from the Arthur engineering team on PRs that modify sensitive paths across released containers, deployment infrastructure, CI/CD, database migrations, and credential handling (#2421)
  • CI workflows now authenticate container and service image pulls to Docker Hub, resolving rate-limit failures during container initialization (#2435)
  • Updated the isort dependency to v9, bringing improved lazy import handling, better performance, and Python 3.15 compatibility (#2198)
  • Updated the sentence-transformers dependency to v6.1.0, with enhanced multimodal input handling and fixes to loss formulas (#2170)

These changes strengthen security governance around critical code paths and improve the reliability and currency of the build and dependency ecosystem.

Internal & Automation
  • Production release version increment consolidating accumulated changes (#2422)
  • Production version bump to 2.1.995 triggering the build and deployment pipeline (#2434)

Release notes generated by Louisa

Assets 2
Loading