Blog

SUPERWISE Sentinel 1.39.0: Output token savings, two new guardrails and a test sandbox

Sentinel 1.39.0 adds Caveman output token optimization, token-saving reports, Language check and Image resolution guardrails, and a guardrail test sandbox.

SUPERWISE®Published

Share
Back to Blog

SUPERWISE® Sentinel 1.39.0 is live. This release is about getting more out of every AI request: cutting token cost, catching more at the guardrail layer, and making the value visible in your dashboards.

Version 1.39.0 is now available. Here’s what’s in it.

Token optimization

  • Caveman: output token optimization. A new option in a Sentinel’s Token optimization tab that keeps the model’s responses terse. Code, numbers, identifiers and error strings stay intact; the filler goes. Choose an intensity of Lite, Full or Ultra. It’s off by default, and an admin turns it on for each Sentinel.
  • Token-saving telemetry and reports. Sentinel now measures what optimization saves and shows it across the app: a Token saving section on the FinOps and Operational dashboards, a stat on the Home page, and per-interaction savings in Live activity. Headroom savings are measured; Caveman savings are estimated.

Guardrails

  • New guardrail: Language check. Detects a prompt’s language and blocks requests that aren’t in your allowed languages. Set an allow or block list and a confidence threshold, which is 0.7 by default.
  • New guardrail: Image resolution. Checks the dimensions of images attached to a prompt and blocks those above a maximum or below a minimum. It’s Sentinel’s first image-based guardrail.
  • Guardrail Test sandbox. Try a guardrail’s staged settings before you publish them. Give it a sample request, or an image for Image resolution, and see whether it Passes, Would remediate or Would block. The test runs in the control plane, with no effect on live traffic or telemetry.

See the docs for Language check, Image resolution and testing a guardrail before you publish.

Sentinel CLI

  • Claude Desktop support on macOS and Linux. sentinel proxy config apps now includes claude_desktop. With it enabled, sentinel proxy on points the Claude desktop app at your gateway. It needs the Anthropic provider and a one-time browser sign-in to store a token.
  • Gateway status. sentinel gateway status shows whether Docker and the container are running, the gateway URL and its health, the running version and whether an update is available, the Sentinel ID, and a link to the Sentinel in the app.
  • Gateway logs. sentinel gateway logs prints the local gateway container’s logs. --tail sets how many lines, 50 by default.
  • Configuration moved under proxy. The configuration commands are now sentinel proxy config url, providers and apps, with a non-interactive one-liner using --url, --providers and --apps.

Also in this release

  • Trial enforcement. Trial status and an end-of-trial block flow across the app.
  • Accessibility. WCAG 2.x level AA conformance improvements across the app.

Get started

New to Sentinel? See what it does to a real prompt in 30 seconds, with no install and no sign-up, or get Sentinel for Windows. Every Sentinel self-service tier starts with 30 days free.

Read the full 1.39.0 release notes and the Sentinel documentation.

Next step

See what Sentinel checks, where it runs, and what each plan includes. See SUPERWISE Sentinel.