VibeAround v0.7.17: one reliable turn lifecycle across Web, TUI, and IM

VibeAround v0.7.17 is out.

This release focuses on a deceptively hard problem: when the same local coding-agent session is accessible from Web, TUI, and messaging channels, every surface needs to agree on what “active,” “queued,” “stopped,” and “finished” mean.

What changed:

• Web, TUI, and IM now share one turn lifecycle. A conversation is not considered idle until its complete queue has drained.

• Stop and cancel wait for the underlying agent to settle, with a bounded forced-shutdown path when it does not.

• History, activity, pending messages, permissions, and session state stay isolated within the correct conversation route.

• Concurrent settings, profile, workspace, onboarding, and launcher changes preserve unrelated updates.

• Managed channel plugins are installed from pinned revisions through staging, so failed preparation leaves the current installation intact.

• Each channel instance gets isolated runtime state and cache directories, while onboarding follows the authentication method declared by the plugin.

• Active agent threads use a soft-limited warm pool, preserving continuity without retaining every thread indefinitely.

Agent profiles can also carry their own API type, base URL, headers, and model list, making provider routing more explicit for each agent and workflow.

The direction remains local-first: agents, repositories, credentials, and execution stay on your machine. VibeAround provides the continuity and control layer around the agents you already use.

Release and downloads:

https://github.com/jazzenchen/VibeAround/releases/tag/v0.7.17

GitHub:

https://github.com/jazzenchen/VibeAround

If you use coding agents across desktop, browser, terminal, or messaging, where does continuity usually break for you: queued follow-ups, cancellation, route isolation, or configuration changes?