Skip to content

2.1.717

Choose a tag to compare

View all tags
@arthur-louisa arthur-louisa released this 20 Jul 21:04
2.1.717
666f7bb
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

🚀 Arthur Engine Release

July 20, 2026

This release brings multimodal prompt support and richer evaluation reporting to the Prompts Playground and Agent Experiments, alongside a wave of UI fixes, expanded Helm deployment documentation, and important security patches.

Prompts & Experiments

Prompts Playground

  • Added image attachment support to user-role messages, enabling multimodal prompts that survive edits and page reloads via a new "Add image" button and thumbnail UI (#1902)
  • Refined prompt content in the task tour feature to improve clarity and onboarding guidance (#1887)

Agent Experiments

  • Agent Experiments now display evaluation results and status on the summary page, including an Overall Eval Performance section with pass-rate cards and a per-test-case Evals column with color-coded status chips (#1834)
  • Fixed Task-card Success % for continuous evaluations to exclude non-terminal statuses (SKIPPED, PENDING, RUNNING) from the denominator, so success rates are no longer artificially deflated (#1960)

The Prompts Playground now supports composing multimodal prompts with images, while Agent Experiments reach parity with Prompt Runs by surfacing evaluation outcomes and accurate pass rates directly on the summary page.

User Interface

Datasets

  • Fixed the "Add to Dataset" wizard so the Cancel button correctly closes the dialog (#1933)
  • Resolved a crash in add-to-dataset span search when span values are non-string types, safely handling JSON objects and numeric fields (#1931)
  • Fixed dataset search so records can now be retrieved by their Row ID (#1940)

Observe & Evaluators

  • Repositioned the AI Assistant button so it no longer overlaps pagination controls in the Observe tab (#1939)
  • Corrected overlapping "Last Updated" and "Actions" columns in the evaluators list (#1935)

Settings

  • Added four Australian timezones (Brisbane, Adelaide, Darwin, Perth) to the Settings timezone selector, covering Eastern, Central, and Western zones (#1932)

A series of layout and interaction fixes across Datasets, Observe, and Evaluators improve everyday usability, and Australian users gain full regional timezone coverage.

Deployment & Helm Charts

Configuration Fixes

  • Added the missing genaiEngineSecretStoreKeyName field to the umbrella values.yaml.template, resolving Helm install/upgrade schema validation failures (#1922)
  • Added auditLog, gcpDiscoveryConfig, and genaiEngineOpenAIVerifySSL keys to the umbrella template, preventing nil-pointer dereference errors and worker boot crashes on fresh installs (#1924)

Documentation

  • Added a "Using a private image repository" section to all three Helm chart READMEs, with step-by-step guidance for mirroring images and configuring pull secrets (#1930)
  • Documented ingress TLS options and HTTPS guardrails, including topology-aware config templates, a conditional TLS block, and post-install warnings for common misconfigurations (#1938)
  • Clarified that genaiEngineInternalIngressHost must include the https:// scheme, addressing a URL-format documentation gap (#1941)

Umbrella template fixes ensure fresh Helm installations succeed out of the box, while expanded documentation makes private-registry deployments and HTTPS configuration far clearer for operators.

Security

  • Updated axios to v1.18.0 to fix a critical denial-of-service vulnerability (GHSA-pmv8-rq9r-6j72) in FormData field-name parsing (#1964)
  • Updated pytest to v9.0.3 to address CVE-2025-71176, a tmpdir vulnerability on UNIX systems; note that pytest v9 introduces breaking changes requiring a pytest-asyncio upgrade (#1830)

These security updates protect against a critical DoS vector in axios and a local privilege-escalation vulnerability in pytest.

Dependencies & Maintenance

Tooling & Libraries

  • Updated sentencepiece to v0.2.2, bringing faster tokenization, native NumPy support, type stubs, and parallel encoding APIs (#1947)
  • Updated tailwindcss monorepo to v4.3.3 with improved file watching, color theme matching, and CSS nesting support (#1957)
  • Updated swagger-typescript-api to v13.12.5, fixing crashes on boolean JSON schemas (#1949)
  • Updated the eslint monorepo to v9.39.5 with a require-cache reliability fix (#1950)
  • Updated @floating-ui/react to v0.27.20 for improved scroll listener performance and TypeScript support (#1926)
  • Updated @tanstack/react-form to v1.33.1, re-exporting useSelector from TanStack Store adapters (#1921)
  • Updated @tanstack/react-form to v1.33.2 for improved form field handling on name changes (#1928)
  • Updated langchain to v1.3.13 (#1929) and v1.3.14 (#1956)
  • Updated arthur-client across genai-engine and ml-engine to v1.4.2261 (#1927), v1.4.2262 (#1944), v1.4.2265 (#1953), and v1.4.2267 (#1955)

Analytics

  • Updated @amplitude/analytics-browser to v2.44.6 (#1943), v2.45.0 with event property attribution enabled by default (#1951), v2.45.1 (#1952), and v2.45.2 (#1954)
  • Updated @amplitude/plugin-session-replay-browser to v1.33.1 (#1920) and v1.33.3 (#1948)

Internal

  • Restored code review functionality by adding arthur-blazity-ai-workflow to the allowed bots list (#1945)

Routine dependency, tooling, and analytics updates keep the engine current with the latest upstream improvements and stability fixes.

Release notes generated by Louisa

Assets 2
Loading