Release notes
Open SWE desktop-v0.3.9
What's Changed
- fix(slack): initialize reply guard in remote runs by @ramon-langchain in https://github.com/langchain-ai/open-swe/pull/4400
- feat(ui): let a person open the thread's browser before the agent does by @ramon-langchain in https://github.com/langchain-ai/open-swe/pull/4364
- chore(deps): refresh LangChain and LangGraph minimums by @open-swe[bot] in https://github.com/langchain-ai/open-swe/pull/4395
- feat: track user and repository usage with PostHog instead of Segment by @ramon-langchain in https://github.com/langchain-ai/open-swe/pull/3888
- feat(threads): make pre-MDA threads read-only and move new tags to MDA by @ramon-langchain in https://github.com/langchain-ai/open-swe/pull/4393
- fix(ui): play GitHub attachment videos inline by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4405
- fix(human-review): enforce work hours for agent-selected reviewers by @ramon-langchain in https://github.com/langchain-ai/open-swe/pull/4392
- fix(reviews): keep note drafts with their note and scope page state by lifetime by @ramon-langchain in https://github.com/langchain-ai/open-swe/pull/4294
- fix(slack): only notify when native Stop fails by @ramon-langchain in https://github.com/langchain-ai/open-swe/pull/4357
- perf(reviews): load the review diff without a compare call or serial pages by @ramon-langchain in https://github.com/langchain-ai/open-swe/pull/4296
- build(mda): vendor the MDA SDK's mda/0.10 branch for the Linear and GitHub channels by @christian-bromann in https://github.com/langchain-ai/open-swe/pull/4305
- feat(linear): receive Linear through the MDA deployment's Linear channel by @christian-bromann in https://github.com/langchain-ai/open-swe/pull/4187
- feat(github): start issue mentions through the MDA deployment's GitHub channel by @christian-bromann in https://github.com/langchain-ai/open-swe/pull/4191
- chore(deps): remove speculative third-party upper bounds by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4403
- feat(mda): replace
spawn_workerwith MDA async subagent workers by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4267 - feat(mda): run file and shell tools in the sandbox from MDA by @ramon-langchain in https://github.com/langchain-ai/open-swe/pull/4355
- ci: show a superseded run's E2E gate as cancelled, not failed by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4413
- feat(cli): wait for task workers before
oswe runexits by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4273 - ci: cache Chromium's system packages and retry stalled apt downloads by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4414
- feat(ui): show each worker launch as a chip in the conversation by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4308
- feat(mda): settle launched workers from the completion webhook by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4382
- fix(slack): validate breakout destination channel IDs by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4415
- chore(deps-dev): bump the npm_and_yarn group across 2 directories with 1 update by @dependabot[bot] in https://github.com/langchain-ai/open-swe/pull/4411
- fix(tracing): preserve destinations when backfilling git metadata by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4412
- ci(mda): forward Linear app credentials to mda deploy by @ramon-langchain in https://github.com/langchain-ai/open-swe/pull/4418
- fix(remote-runtime): name backend hook trace spans by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4417
- feat(merge-queue): dispatch and track candidate validation by @keshivtandon in https://github.com/langchain-ai/open-swe/pull/4366
- fix(ui): show keyboard hints on shortcut-enabled submit buttons by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4404
- fix(github): retry organization membership checks by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4377
- fix(runtime): use Agent Server dev build for cancellation recovery by @mdrxy in https://github.com/langchain-ai/open-swe/pull/4419
- fix(review): allow approval policies up to 15k characters by @bwhiting2356 in https://github.com/langchain-ai/open-swe/pull/4408
- chore(desktop): bump version to 0.3.9 by @johannes117 in https://github.com/langchain-ai/open-swe/pull/4420
Full Changelog: https://github.com/langchain-ai/open-swe/compare/desktop-v0.3.8-nightly.20261022045508...desktop-v0.3.9