Skip to content

v0.17.0

Latest

Choose a tag to compare

@github-actions github-actions released this 28 Sep 09:53
· 16 commits to main since this release
995775a

What's Changed

  • docs: add v0.16 release notes to the README by @mrwadams in #91
  • ci: drop Python 3.10 in favour of a 3.11 floor by @mrwadams in #95
  • Keep the elapsed timer ticking during long model-streaming waits by @mrwadams in #92
  • Degraded success and retry for optional enrichment by @mrwadams in #93
  • Accessibility and browser-console hygiene by @mrwadams in #94
  • Run the browser checks in CI, and declare tenacity by @mrwadams in #96
  • Harden the scenario-generation journey: persistent state, guided setup, and resilient progress by @mrwadams in #97
  • Trim the verbose UI copy across the app by @mrwadams in #100
  • Pin feedback to the result's own run, with no session fallback by @mrwadams in #102
  • Refresh the model registry with the latest models from Anthropic, OpenAI, Google, Mistral and Groq by @mrwadams in #101
  • Follow-ups from verifying #97: feedback run id, misreported technique-lookup failure, upstream sidebar label by @mrwadams in #103
  • Custom Scenarios hides the technique picker and blames the user when techniques fail to load by @mrwadams in #107
  • Make the technique-lookup failure blocker actionable by @mrwadams in #108
  • Threat Group page: empty-result and lookup-failure branches are pre-empted by the dropdown filter by @mrwadams in #109
  • Make the Custom Scenarios technique-load blocker actionable by @mrwadams in #113
  • deps(deps): update streamlit requirement from >=1.55 to >=1.63.0 by @dependabot[bot] in #116
  • deps(deps): bump actions/cache from 4 to 6 by @dependabot[bot] in #117
  • Deepen the MITRE resolution seam: one kill-chain resolver, one technique-ID form, one defence entry point by @mrwadams in #114
  • One AI-insider selection rule for the page and the MCP server by @mrwadams in #115
  • deps(deps): update streamlit requirement from >=1.63.0 to >=1.64.0 by @dependabot[bot] in #119
  • Delete core/readiness.py and the dead is_ready parameter by @mrwadams in #124
  • Make the Assistant conversation a module keyed by scenario identity by @mrwadams in #123
  • AttackGen Assistant: apply refinements to the downloadable files, not just the chat by @mrwadams in #118
  • Show the whole Assistant chat history after a rerun by @mrwadams in #125
  • Repin the python:3.14-slim base image to the current digest by @mrwadams in #126
  • Add Claude Opus 5.5 and make it the Anthropic default by @mrwadams in #128
  • release: bump version to 0.17.0 by @mrwadams in #127

Full Changelog: v0.16.0...v0.17.0