AddedAdded file-tree context menu actions to add notes, folders, PDFs, and sidebar files directly to the current chat composer as context pills
AddedAdded Add to New Chat and Add Selected to New Chat actions from the file tree, opening a fresh agent chat before attaching the chosen context
AddedAdded multi-selection-aware chat context actions for selected notes and sidebar files in the file tree
AddedAdded local app diagnostics logs for Electron main, renderer, and native backend events, with documented log locations and privacy notes
AddedAdded saved-chat crash recovery docs and recovery flow for AI conversations stored locally under each vault's .neverwrite/sessions directory
AddedAdded native saved-session resume support for Codex when available, with a saved-transcript fallback path when direct runtime resume is unavailable or fails
ChangedChanged file-tree path copy actions to consistently use Copy Full Path and copy absolute paths for notes, folders, PDFs, and sidebar files
FixedFixed external vault refresh handling so ambiguous external deletes, including folders that look like Markdown notes, refresh the vault structure instead of leaving the file tree stale
FixedFixed closing active AI agent tabs so NeverWrite asks for confirmation consistently, including Cmd/Ctrl+W and multi-tab close paths
FixedFixed Codex saved chats so restored, detached, resumed, or crash-recovered sessions keep enough transcript context to continue without losing the prior conversation
FixedFixed Codex subagent breadcrumb Open actions so they continue working after restore or resume even when live session identifiers differ from saved history
FixedFixed detached windows and tab reattachment for AI chats so related parent, child, and sibling subagent sessions transfer and hydrate together
FixedFixed file-tree chat context targeting so context menu actions and drops attach to the intended chat, including newly opened chats whose composer is still mounting
FixedFixed normal native backend shutdown so expected sidecar exits no longer appear as app errors
SecurityHardened diagnostic log redaction for prompt, transcript, message, content, token, secret, authorization, and API key fields
