Skip to content

v1.4.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 12 Oct 15:17
· 601 commits to main since this release
v1.4.0
00fc3b5
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

v1.4.0

This release adds explicit Plan, Red, and Blue run modes to the OSS launcher, CLI, and web dashboard. Red can start from one confirmed target and authorization reference; Soundwave's eight-document interview is now an optional Plan mode. It also brings the current production LiteLLM model routes into OSS and makes all nine planning documents visible in the Plan view.

Changes

  • Plan, Red, and Blue modes across the launcher, CLI /mode command, and web live view. Mode-specific sessions preserve their threads; Red rechecks RoE approval and Blue checks both sensor and monitor health (#919).
  • Current production LiteLLM model routes for OSS agents and CLI (#916).
  • Five additional Plan document views (#893).
  • Selectable local Ollama model ID in the CLI (#905).
  • Complete LiteLLM model count and concise service health details (#897).
  • Settings collection error handling, including non-JSON HTTP failures (#895).
  • Engagement-scoped report tool calls (#907).
  • Web unit tests in CI and a maintained Unreleased changelog section (#917).
  • First-screen English and Korean README coverage of Plan/Red/Blue, REA/Ghidra reversing, Blue Cell, specialist agents, and review surfaces (#918).

Blue Cell local sensing and the resident defensive agent were introduced in v1.2.2. Blue mode now brings that agent into the same explicit mode flow. REA/Ghidra reversing remains available from v1.3.0 and v1.3.1; the README now explains when the Reverser starts it and where the original Evidence is retained.

Verification

Before this mode PR, the integrated source passed make quality: 5,492 Python tests passed (43 skipped), 69 CLI tests passed, 14 web tests passed, and the web production build completed. The mode branch passed 77 CLI tests, 17 web tests, the Go launcher test suite, CLI/web TypeScript checks, and the web production build. A running build connected to a 37-model LiteLLM fixture returned modelCount: 37 and 37 models loaded from /api/health.

See the full changelog.

Assets 20
Loading