Skip to content

feat: desktop computer-use APIs with neko streaming#264

Merged
NathanFlurry merged 4 commits into
mainfrom
desktop-computer-use-neko
Mar 17, 2026
Merged

feat: desktop computer-use APIs with neko streaming#264
NathanFlurry merged 4 commits into
mainfrom
desktop-computer-use-neko

Conversation

@NathanFlurry

Copy link
Copy Markdown
Member

Summary

  • Add desktop runtime management (Xvfb, openbox, dbus) with start/stop/status APIs
  • Screen capture, mouse/keyboard input, and scroll APIs
  • Video streaming via neko binary extracted from m1k1o/neko container
  • Desktop recording support (start/stop/list/download)
  • TypeScript SDK desktop support with DesktopStream class
  • React DesktopViewer component
  • Inspector Desktop tab with live view, screenshots, recording
  • Docker test rig for integration testing

Test plan

  • Docker build succeeds
  • Desktop start/stop lifecycle works
  • Screenshot capture returns valid PNG
  • Mouse click and keyboard type dispatch correctly
  • Neko video stream connects in browser
  • Recording start/stop produces downloadable video

🤖 Generated with Claude Code

@railway-app
railway-app Bot temporarily deployed to sandbox-agent / sandbox-agent-pr-264 March 17, 2026 00:57 Destroyed
@railway-app

railway-app Bot commented Mar 17, 2026

Copy link
Copy Markdown

🚅 Deployed to the sandbox-agent-pr-264 environment in sandbox-agent

Service Status Web Updated (UTC)
website ◻️ Removed (View Logs) Web Mar 17, 2026 at 9:36 am
foundry-mock ◻️ Removed (View Logs) Mar 17, 2026 at 9:36 am

@railway-app
railway-app Bot temporarily deployed to sandbox-agent / sandbox-agent-pr-264 March 17, 2026 06:59 Destroyed
@railway-app
railway-app Bot temporarily deployed to sandbox-agent / sandbox-agent-pr-264 March 17, 2026 09:36 Destroyed
NathanFlurry and others added 4 commits March 17, 2026 02:36
Add desktop runtime management (Xvfb, openbox, dbus), screen capture,
mouse/keyboard input, and video streaming via neko binary extracted
from the m1k1o/neko container. Includes Docker test rig, TypeScript SDK
desktop support, and inspector Desktop tab.

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Improve desktop streaming architecture, add inspector dev tooling,
React DesktopViewer updates, and computer-use documentation.

Co-Authored-By: Claude Opus 4.6 <[email protected]>
…treaming

Adds desktop computer-use endpoints (windows, screenshots, mouse/keyboard,
launch/open), enhances neko-based streaming integration, updates inspector
UI with desktop debug tab, and adds common software test infrastructure.

Co-Authored-By: Claude Opus 4.6 <[email protected]>
@NathanFlurry
NathanFlurry force-pushed the desktop-computer-use-neko branch from 944309c to dff7614 Compare March 17, 2026 09:36
@railway-app
railway-app Bot temporarily deployed to sandbox-agent / sandbox-agent-pr-264 March 17, 2026 09:36 Destroyed
@NathanFlurry
NathanFlurry merged commit 3b8c745 into main Mar 17, 2026
2 of 6 checks passed
NathanFlurry added a commit that referenced this pull request Mar 17, 2026
- Fix connectDesktopStream silently dropping RTCPeerConnection and rtcConfig options (client.ts)
- Fix DesktopViewer useEffect dependency causing reconnect loop (store callbacks in refs)
- Fix TOCTOU race condition in DesktopRecordingManager::start() (merge lock scope)
- Fix incomplete cursor bounds check in composite_cursor_region (add right/bottom checks)
- Add DesktopViewer to react-components.mdx documentation
- Remove hardcoded visual styles from DesktopViewer (make unstyled by default per sdks/CLAUDE.md)
- Export DesktopViewerClassNames type for consumer styling

Co-Authored-By: Claude Haiku 4.5 <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant