-Added a renderer for documents downloaded in chats. -Added export/share view for non-rendered file types. -Fixed a bug where copying text from chats with markdown formats only copied the final element. -Improved setup wizard and onboarding hints for dashvox cloud.

- Settings cleaned up and reorganized: clearer info buttons, updated descriptions, and better grouping so related settings live together. - SMS settings moved higher up for cloud users. - Links in chat are now tappable. - Optimizations to session auto-refresh. - Longer silence gap in driving mode, so a brief pause mid-sentence no longer cuts you off. - CarPlay now loads your chats and messages on its own, without needing to open the phone app first.

1. Auto-send now works in noisy environments Hands-free driving mode wasn't sending messages in rooms with steady background noise (e.g. office AC). The old detector used a fixed mic-energy threshold that constant noise held open, so the "you stopped talking" timer never fired. Now it endpoints on transcript-growth stall so it sends when the recognizer stops producing new words, which ambient noise can't spoof. 2. Spoken notifications outside the app (Communication Notifications) Enabled the Communication Notifications entitlement. When DashVox is closed or backgrounded and a reply notification arrives, iOS reads it aloud over CarPlay/AirPods, with Focus bypass and instant Watch delivery. (Requires the backend to send mutable-content:1 and enabling Settings → Notifications → DashVox → Announce Notifications.) 3. Terminal commands now work on backend-managed sessions send_terminal_command was returning "[No direct SSH session on this device]" for backend-managed sessions whenever the app wasn't holding a live on-device SSH connection. It now forwards those commands to the backend as a raw shell command and returns the output.

- Added a hands-free accessible driving chat ui feature that emulates CarPlay for users who do not have CarPlay and prop their phone on the dash to use DashVox features. - Fixed TTS bug where previous responses are repeated before the current one. - Introducing DashVox Ultimate: a subscriber cloud tier that provides a managed VM service + pre-configured agents.

-CarPlay Speak dead-mic: Speak no longer has a chance of closing after "Preparing". The audio engine now rebuilds against the live mic route (the old engine was stuck on a stale 0 Hz format from launch), with an automatic fallback to the iPhone mic when the car's CarAudio route has no usable input. -TTS cutting off at less than or greater than signs: The text-to-speech read used to go silent the moment it hit a less than or greater than sign. Angle brackets are now stripped so the full response reads aloud. -TTS reading the shell prompt: spoken session replies no longer lead with the terminal prompt (e.g. "user@DESKTOP..."). -Stale lists after switching backends: switching between cloud and self-hosted now wipes and reloads the chat and session lists immediately, instead of showing the previous backend's data until a manual refresh.

-Fixed session chat UI not loading at the bottom of the chat -Fixed CarPlay bug where music wouldn't properly unpause after TTS/Recording -Fixed and refined CarPlay direct session chat -Further improvements and Enhancements

Provider-agnostic LLM support - DashVox can now connect to any LLM provider: OpenAI, Anthropic, Gemini, or a fully custom endpoint instead of only built-in ones. - Providers are configured through a request/response descriptor: you describe the API's request body and where the reply lives, and DashVox adapts to it. - Family-aware prefills (OpenAI/Anthropic/Gemini) give you a known-good starting template you can tweak, rather than a blank box. Descriptor editor - Dedicated request/response builder with insertable, color-coded variable tokens, replacing the old raw-JSON field. - Tools editor: tool-definition and tool-result templates with token palettes, the five tool-call response paths, OpenAI-shaped prefill, and the tools section now shown automatically when a prefill uses tools. - Added keyboard dismissal (drag-to-dismiss + a "Done" bar) and equalized palette chip heights. Session chat UI - New chat-bubble interface for agent sessions: full turn history rendered as bubbles, parallel to the conversation chat, scoped to one code session. - Agent / Terminal mode toggle, with terminal turns shown in distinct monospaced bubbles. - Scroll-up pagination through older turns, inline attachments (images/PDFs), pinch-to-zoom and wrap-text controls for terminal output. - Live updates on send/response, plus refresh on foreground and notification arrival; swipe-down to dismiss the keyboard. - A session's detail view auto-opens the moment you create it. Session chat queuing & notifications - Session-command queuing (backend 1.0.111): follow-ups sent while a session is RUNNING are parked and shown as cancelable pending bubbles on phone + watch, draining automatically when the round responds. - USER-started session-completion banners now deep-link into the session chat on iPhone and Watch; the redundant banner is suppressed while you're viewing that session, and its response is pulled in immediately. - Sessions show a spinner while RUNNING (list + detail). Watch - Restructured navigation: a single NavigationStack outside the TabView so a pushed chat/session detail covers the pager (no more accidental page-swipe inside a detail). - Added a notification-tap handler routing chat vs. session banners correctly. - Session detail input fix: the mic button is now reliably visible (moved to a safeAreaInset). - Chat-list scroll indicator no longer lingers into the chat detail view. Voice & TTS - Session voice dictation now matches the chat's tuned live-transcription flow.

- Fixed: Apple Watch dropping the phone connection off WiFi. When the Watch had no WiFi of its own, brief Bluetooth hiccups (phone in a pocket or console, a moving car) would instantly show "reopen DashVox to reconnect" — even though the phone was fine. The Watch now waits up to 30 seconds for the link to recover before reporting a problem, so momentary drops no longer interrupt you. - Clarified when the phone needs to be open: the Watch can relay through your iPhone even while the DashVox app is suspended or backgrounded. You only need to reopen it if you've fully force-quit it (swiped it away in the app switcher). - Chats with photos open almost instantly and use far less data. Full images now download only when you tap them. Image-heavy chats that used to take seconds are near-instant. - Smoother typing in chats with images. The keyboard no longer lags or stutters while typing in conversations that have photos in them. - Fixed open-chat-through-notification bug where chat briefly displaying stale content before catching up. Tapping a notification now opens the up-to-date chat. - Images sent by the assistant show up on your watch, with a small preview thumbnail in the chat. - Tap an image to open it full-screen: turn the Digital Crown to zoom in and drag to pan. Zoomed images now stay properly inside the frame. - Replies now read aloud through your phone and AirPods when a session you dispatched finishes even with your watch and phone pocketed and screens off so long as both apps aren't suspended. - Opening a chat no longer re-dictates the unseen message you're already looking at. - Fixed an occasional double read-aloud of the same reply. - Images load far faster and use a fraction of the data. Full images now download only when you tap them. - Smoother scrolling back through long conversations. - Reduced the brief freeze the first time you open the keyboard. - Fixed a rare glitch where the chat input bar could jump to the middle of the screen instead of staying at the bottom.

- Fixed: CarPlay voice replies could go silent mid-drive and never recover. When a navigation app (e.g. Google Maps) spoke a turn-by-turn prompt, it interrupted the audio session and left text-to-speech stuck — after that, nothing was read aloud, including "Read Last Response." TTS now detects these interruptions and automatically restores itself, so replies keep reading the whole drive. - Added: a "message sent" sound so you get audible confirmation your dictated message went through without looking at the screen. - Added: a "Last reply" row at the top of the CarPlay chat screen showing the agent's most recent reply as text — a glanceable backup for when audio is interrupted. Tap it to hear the reply again (this also re-arms TTS if it was interrupted). - Improved: more resilient audio handling under heavy navigation, with an automatic retry so TTS recovers cleanly instead of staying broken.

-Updated search keywords so the app appears in app store search -Fixed background ("Run Anywhere") sessions reporting "[No SSH connection]". The agent now reads their live output from the server instead of the empty on-device buffer, and SSH errors now say whether the session is missing or the connection just dropped -Fixed email verification not sending

-Added full Siri functionality -Fixed a double notification bug -Added Twilio SMS messaging support -Added support for http routing through phone when no connection is available on watch. -Added auto chat compacting to prevent context and token use excess -Added onboarding/setup wizard for new users -Added image/file uploading for agent chat -Added auto chat compacting and context optimization -Fixed CarPlay Functionality -Rearchitected the chat ui on phone and watch to minimize latency on large/long running chats. -Further optimizations and enhancements