v0.18.0 — Conversational work across web and mobile

Optio v0.18.0 brings conversational work creation to the web and mobile apps.

• Ask Optio to discover repositories and machines, propose work, and create it after approval through the Optio controller MCP.
• The assistant is a saved waits-for-me session using Claude Code, Codex, Copilot, Gemini, OpenCode, or Cursor. Choose its harness and model in Settings with the same controls as work creation.
• Float or dock the assistant, dictate an editable message in supported browsers, and continue the saved conversation from its session. The compact header shows connection status and the selected harness/model.
• iOS and Android have a dedicated Optio Agent tab with conversation history, approval cards, message queues, and harness/model settings. Insights remains in More. Keyboard dictation works through the device keyboard.
• Pod sessions now open at /work/:id. Existing /sessions/:id links redirect there while preserving terminal panes, and the session sidebar links back to All work.
• Send while the agent is working: messages queue durably across clients and run in order. Remove a queued message before it starts.
• A settings icon opens the agent configuration directly; unsupported browser dictation is disabled with an explanatory tooltip.
• Refreshed account settings, shorter message composer, and clearer minimize, clear-conversation, and open-session controls.

This release includes a database migration for session harness configuration and durable tool approvals and queued messages; migrations run during API startup. Browser dictation depends on browser speech-recognition support; Firefox users can use operating-system keyboard dictation. Native client source is included; this release does not itself publish App Store, TestFlight, or Play Store builds.