← News
Release

qm v0.1.15

Memory as one row per fact — semantic recall across authorized notebooks, plus a nightly tidy (#2229, #1129, #2323, #2232). Goals that handle getting stuck — a governor steps back, then pauses;…

Originally published on github.com →

Highlights

  • Memory as one row per fact — semantic recall across authorized notebooks, plus a nightly tidy (#2229, #1129, #2323, #2232).
  • Goals that handle getting stuck — a governor steps back, then pauses; people can resume or cancel (#2132, #2116, #2271).
  • Reply to Slack conversations from the web UI, with unread dots for new agent replies (#2383, #2250).
  • Smoother app connections — missing connections offered inline and continued after consent; per-user OAuth for admin-configured MCP servers (#2043, #1812).
  • Admin decision rules that run on every turn and can be enforced from chat (#2192, #2334, #2336).
  • Deploy resilience — retry-safe tool calls re-run after a deploy interrupt; people's turns run before background work (#2055, #2356).
  • New options — DigitalOcean Managed Agents and E2B computer backends, Browser Use Cloud for browsing, Claude Haiku 5.5, GPT-6.1 Sol Ultrafast (#2128, #2370, #2367, #2109, #2143).
  • Breaking: Codex and OpenCode harnesses removed; session_tape dropped after migrating its data (#2277, #2273, #2281).

What's Changed

New Contributors

Full Changelog: https://github.com/yc-software/qm/compare/v0.1.14...v0.1.15