docs: update README, CHANGELOG, and release workflow for v1.3.0

- Add Tauri v2 to tech stack, standalone desktop app to features
- Document persistent Whisper server, ComfyUI deep-scan fixes
- Update release workflow default tag to v1.3.0

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: PurpleDoubleD

.github/workflows/release.yml

@@ -8,7 +8,7 @@ on:
       tag:
         description: 'Release tag (e.g. v1.2.1)'
         required: true
-        default: 'v1.2.1'
+        default: 'v1.3.0'
 
 permissions:
   contents: write

CHANGELOG.md

@@ -11,8 +11,14 @@ All notable changes to Locally Uncensored are documented here.
   - Ollama context window warning when model has insufficient context
   - Automatic embedding model download (nomic-embed-text)
   - Per-conversation RAG toggle and source citations
+- **Standalone Desktop App**: Full Tauri v2 Rust backend — .exe runs without Node.js or dev server
+  - 15 Rust commands replacing Vite middleware (process management, downloads, search, agents, voice)
+  - Frontend auto-detects Tauri vs browser and routes accordingly
+  - Ollama, ComfyUI, and Whisper auto-start on app launch
+  - Clean process shutdown on app exit
 - **Voice Integration**: Talk to your AI and hear responses
-  - Push-to-talk microphone button with Web Speech API
+  - Persistent Whisper server loads model once (~2.5 min), then transcribes in ~2s
+  - Push-to-talk microphone button with local faster-whisper (100% offline, no cloud)
   - Text-to-speech on any assistant message with sentence-level streaming
   - Voice settings (voice selection, rate, pitch)
   - Auto-send transcribed text option
@@ -26,6 +32,9 @@ All notable changes to Locally Uncensored are documented here.
 ### Fixed
 - Cross-platform Python detection for code execution (Windows Store alias handling)
 - Web search now falls back to Brave Search when DuckDuckGo returns CAPTCHA
+- ComfyUI auto-discovery now scans up to 4 levels deep (finds nested installs with spaces in path)
+- Ollama/ComfyUI spawn no longer opens extra console windows on Windows
+- Whisper transcription no longer times out (was re-loading 145MB model on every request)
 
 
 ## [1.0.2] - 2026-03-25

README.md

@@ -73,13 +73,15 @@ Tired of switching between Ollama for chat, ComfyUI for images, and another tool
 - **Document Chat (RAG)** — Upload PDFs, DOCX, or TXT files and chat with your documents. Hybrid search with confidence scores.
 - **Voice Chat** — Talk to your AI with push-to-talk and hear responses with text-to-speech. Sentence-level streaming.
 - **AI Agents** — Give your AI a goal and watch it plan, search the web, read/write files, and execute Python code autonomously.
+- **Standalone Desktop App** — Full Tauri v2 Rust backend. Download the .exe, run it — no terminal, no dev server, no Node.js.
 
 ## Tech Stack
 
+- **Desktop**: Tauri v2 (Rust backend, standalone .exe — no Node.js required)
 - **Frontend**: React 19, TypeScript, Tailwind CSS 4, Framer Motion
 - **State**: Zustand with localStorage persistence
-- **AI Backend**: Ollama (text), ComfyUI (images/video)
-- **Build**: Vite 8
+- **AI Backend**: Ollama (text), ComfyUI (images/video), faster-whisper (voice)
+- **Build**: Vite 8 (dev), Tauri CLI (production)
 
 ---