Release v0.1.90 · andrewtheart/multiterm-workbench

What's changed

New Logs viewer in the status bar (replacing Runtime logs) showing application and bridge diagnostics from the current launch alongside durable records from earlier launches.
Diagnostics are persisted as rotating JSONL files under %LOCALAPPDATA%\MultiTerm\Diagnostics, with new Diagnostics settings for retention days, rotation size, and maximum viewer entries (default 14 days / 10 MB / 5000 entries; 0 disables each limit).
Optional Include Copilot CLI logs aggregation, off by default. When enabled, MultiTerm-launched Copilot sessions get a dedicated directory under Diagnostics\Copilot and each aggregated row is labeled with its terminal title and stable terminal ID. Copilot initial tail (KB) controls how much existing content is read per owned log (default 256 KB; 0 follows only new lines). Copilot processes started outside MultiTerm are not included, and Claude is unaffected.
Diagnostic records drop sensitive named fields and strip credentials, query strings, and fragments from URLs. This is bounded sanitization, not comprehensive secret detection; raw Copilot CLI logs remain unredacted and are not removed by JSONL retention.
Electron now captures bridge stdout/stderr and records bridge spawn, exit, restart, and restart-abandoned events. WebSocket close and error events log detail and settle pending bridge requests as disconnected.

The automatic-title schedule now starts only after a terminal receives meaningful input (typed or pasted text, or commands sent by MultiTerm actions such as Run GitHub Copilot), so an untouched terminal no longer spends an AI request.
New timing modes: Progressive delays (ordered list whose last delay repeats) or Repeating interval (fixed minute interval).
A one-time notice appears immediately before the first automatic title request that is ready to run, letting you change timing mode, continue with the pending request, or turn automatic titles off.
Title suggestions are now recorded with timestamp, terminal title, PID, automatic/manual source, and accepted state. See the five most recent from the editable title (or press Down while editing) and the terminal hamburger flyout; open Title suggestion history from the flyout or command palette to search all retained history by title, suggestion, PID, or outcome, with Show 5 more paging. Retention is configurable via History entries under AI assistants (default 500).

Terminals can now keep multiple process-bound notes: add, edit, and delete individual notes from the Notes flyout, Expand a draft into the full notes and command-queue dialog, and create or delete notes there. When a process ends, each note is recovered separately instead of overwriting another.
Automation steps that open a terminal can now target the active page, a new page, or an existing page by title, with suggestions from current page names.
New Run when workstation is condition allows an automation while Windows is locked, unlocked, or in either state; a mismatched or unavailable state is recorded as skipped.
Delivery modes clarified: Run when ready inserts and presses Enter only once the target shell or Copilot composer is ready, and Stage without Enter (formerly Stage when ready) inserts without Enter. Copilot steps now wait for the empty composer.
Git worktree creation and bring-back/merge operations now report step-by-step progress (validating, snapshotting, creating, importing, verifying, merging, conflict checks, committing, and cleanup).

Fixed Copilot provider detection always reporting "GitHub Copilot is not signed in for this Windows account" even with the Copilot CLI installed and signed in, and refresh never changing the result. The SDK host now uses copilot-cli mode instead of Empty mode so stored OAuth credentials are read; session isolation is unchanged because ambient features stay disabled in SessionConfig and tools remain off.

Version bumped to 0.1.90 in package.json, the Inno Setup installer, and the Visual Studio extension manifest.
New lib/runtime-diagnostics.js and lib/copilot-log-aggregator.js modules, a durable diagnostics store in the PowerShell bridge, and a multiterm:configure-diagnostics Electron IPC channel exposed through preload.js.
Added .github/copilot-instructions.md defining behavior-focused test organization; HELP.md and the performance and security docs were updated for diagnostics and Copilot log aggregation.
New and expanded test coverage, including runtime diagnostics, auto-title, title suggestion history, worktree dialog, automations, Copilot log aggregation, and Git worktree suites.

Download the attached MultiTerm-Setup-0.1.90.exe and run it to install MultiTerm Workbench v0.1.90.

MultiTerm-Setup-0.1.90.exe
- Size: 109686593 bytes
- SHA-256: 0476d4a121f7a9b1f387e7b08c6da15420116c0dad1af43c4b239d674583c260