SpotAI v1.4.4 🚀

The ultra-fast, Raycast-inspired AI Spotlight launcher for Windows.
Access local LLMs (Ollama, LM Studio) or cloud models (OpenAI, Anthropic, OpenRouter, DeepSeek) anywhere with a single shortcut (Alt + Space). Now with RAG Local, Advanced Voice Recognition, CLI Injection, Real Embeddings, Interactive Citations, Smart Suggestions, Immersive Reading Mode, and Conversation Export.

Light theme	Dark theme

---

📝 Release History

Every version's changes — from v1.0.0 to the latest — live in the versioned CHANGELOG.md, generated from git tags with npm run changelog:all and committed automatically by the Release workflow.

---

🔥 Key Features

• ⚡ Instant Spotlight Access (Alt + Space): Summon SpotAI anywhere on your desktop in less than 100ms.
• 📋 Automatic Clipboard Context Capture: Selected text or copied code is automatically attached as context for instant explanation, refactoring, or translation.
• 🦙 Ollama Local & Network Server Support: Full compatibility with Ollama running locally or on a remote server/IP on your local network (e.g. http://192.168.1.100:11434).
• ☁️ Multi-Provider Cloud AI: Native support for OpenAI, Anthropic (Claude), Groq, DeepSeek, plus local LM Studio — and any OpenAI-compatible endpoint (OpenRouter, Mistral, Together, vLLM…) with a custom base URL, API key and model.
• 🎯 Native Auto-Insert (Ctrl + Enter): Insert AI responses directly into your active text editor, code editor, or browser with one click.
• 📐 Dynamic Window Moving & Resizing: Drag the title bar to position SpotAI anywhere, and resize it freely (drag the bottom edge or corner) — the size is remembered between sessions.
• 🔒 Encrypted Storage: Sensitive API keys are stored securely using the Windows Credential Manager.
• ⌨️ Configurable Global Hotkey: The Alt + Space shortcut can be changed to any combo (e.g. Ctrl + Shift + Space) from Settings.
• 🧠 Custom System Prompt: Define your own system prompt applied to every request, or keep SpotAI's built-in default.
• 💬 Multi-turn Chat: Conversations with memory, persisted between sessions, with a one-click "New chat".
• 🚀 Auto-start with Windows: Launch SpotAI automatically when you log in (toggle in Settings → General).
• 🔄 Automatic Updates: SpotAI checks GitHub Releases on startup and installs & restarts in one click (Tauri updater, signed).
• 🎙️ Advanced Voice Input: Hold Alt + V (or press the mic button) to speak in 5 languages (EN/ES/DE/PT/FR) with Whisper.cpp and real embeddings.
• 📚 Local RAG — Ask Your Files: Drag files in and ask questions about your code and documentation with local vector semantic search. Supports PDF, DOCX, TXT, MD, CSV, HTML/XML, JSON, YAML, TOML, code (Rust, Python, JS/TS, Go, Java, C/C++, etc.) and config/log files.
• 💻 CLI Injection (/exec): Generate smart PowerShell/CMD commands with a security confirmation step and direct terminal execution.
• 🔍 Real Embeddings: Uses the nomic-embed-text model via Ollama for accurate semantic representations instead of hash-based fallbacks.
• 📦 Zero Heavy Dependencies: Lightweight native Windows binary built with Tauri v2 and Rust.
• ✨ Interactive Citations: AI responses include clickable references [1], [2] that show exact document excerpts when clicked, with automatic highlighting in the document viewer.
• 💡 Smart Question Suggestions: Automatically generates relevant question chips based on uploaded documents to guide user interaction.
• 📖 Immersive Reading Mode: Split-screen view with simultaneous document viewer and chat for enhanced study experience.
• 📤 Conversation Export: Export complete chat history to Markdown format with citations, references, and document sources included.
• 📊 Active Documents Panel: Visual management of indexed documents with real-time status, file details, and individual deletion controls.
• ⏳ Real-time Indexing Feedback: Progress bars and clear status messages during document processing ("Analyzing structure...", "Extracting text...", "Ready to answer").

---

🎙️ Voice Input (Dictation)

Speak instead of typing: hold Alt + V (or click the microphone button in the toolbar) to record, then release — the recognised text is injected straight into the prompt, ready to send with Enter.

Two transcription engines are available in Settings → General → Speech-to-text engine:

• 🪶 Native (Windows SAPI) — zero setup. Uses the Windows speech recognizer in live dictation mode with short silence timeouts for instant results.
• 🧠 Whisper (whisper.cpp) — more accurate and multilingual (en/es/de/pt/fr). The first time you pick it, a download (binary + model) is offered from Settings; afterwards everything runs locally and offline. You can choose the model size (Tiny / Base / Small) and the recognition language — see Tuning Whisper below. Recordings are also saved as WAV so the Whisper engine can transcribe them.

The Whisper install panel (3 phases)

1 · Before download	2 · During download	3 · Ready
Shows the Download Whisper model button for the selected size (Tiny/Base/Small).	A live progress bar tracks the binary and model download phases.	A green check confirms Whisper model ready with the installed model size.

Tuning Whisper: model size & recognition language

In Settings → Voice you can adjust how Whisper transcribes:

• 🎚️ Recognition language — whisper's automatic detection on short clips often assumes English and produces nonsense on other languages. Pin the spoken language (Auto / English / Español / Deutsch / Português / Français) for reliable dictation; when nothing is pinned, the app defaults to your interface language automatically.
• 📦 Model size — Tiny (~75 MB) is the fast default, while Base (~145 MB) and Small (~466 MB) are noticeably more accurate on non-English speech (at the cost of slower transcription). Switching only downloads the model you pick — the binary is shared, and sizes you've already downloaded switch instantly.

💡 Tip: for dictation in Spanish (or any non-English language), pick Base or Small and pin Español — the Tiny model is fast but often misrecognises words.

💡 Tip: Whisper runs fully offline after the one-time download. The native engine needs the Windows speech language pack for your language.

---

🧪 Testing & CI

• Unit tests: cargo test covers provider parsing, history bounding, message formatting, host validation and credential masking; npm test (Vitest) covers i18n completeness across en/es/de and the action prompt templates.
• Continuous Integration (.github/workflows/ci.yml): typecheck, unit tests and production build on every push/PR.
• Releases (.github/workflows/release.yml): pushing a vX.Y.Z tag builds, signs and publishes the GitHub Release automatically (no draft step) with the installers, .sig signatures, the portable zip and the latest.json updater manifest — zero manual steps. The workflow also regenerates CHANGELOG.md from the full tag history and commits it to main as github-actions[bot]. Requires two repository secrets:
  • TAURI_SIGNING_PRIVATE_KEY — the content of ~/.tauri/spotai.key
  • TAURI_SIGNING_PRIVATE_KEY_PASSWORD — the key password (leave empty when the key has none)

ℹ️ The version is managed in a single place: src-tauri/tauri.conf.json. The frontend reads it at build time, so bumping the version there is enough for the installers and the in-app UI.

---

🎹 Keyboard Shortcuts

Shortcut	Action
Alt + Space	Show / Hide SpotAI overlay (configurable in Settings)
Alt + V	Hold to record voice input, release to transcribe into the prompt
Esc	Close / Hide SpotAI window
Enter	Submit prompt
Shift + Enter	Insert new line in prompt box
↑ / ↓	Navigate prompt history
Ctrl + Enter	Auto-insert the last response into the previous app
Ctrl + ,	Open Settings Modal
/ + ↑/↓ + Enter	Open the slash command palette and pick a command
Ctrl + V (with a screenshot)	Attach an image as context for vision models

/ Commands (type and press Enter to run directly)

Command	Action
/new	Start a new conversation
/theme	Toggle between dark / light theme
/capture	Capture a screen region
/incognito	Toggle incognito mode
/settings	Open the settings modal
/hide	Hide the window
/clear	Clear the current conversation
/explain	Explain selected code or text
/fix	Fix a bug in selected code
/refactor	Refactor selected code
/summarize	Summarize selected text
/translate	Translate the selected text
/improve	Improve writing quality and tone
/comment	Add clear inline documentation
/exec	Generate and execute shell commands with security confirmation
/rag	Open RAG panel to ask questions about your local files

---

🚀 Quick Download & Installation

Download the latest version from GitHub Releases:

• 🗂️ Portable (no install): SpotAI-Portable.zip — unzip and run SpotAI.exe directly, nothing to install (requires Windows 10/11 with WebView2).
• ⚡ Installer (NSIS): SpotAI_1.4.4_x64-setup.exe
• 📦 Installer (MSI): SpotAI_1.4.4_x64_en-US.msi

(All assets — installers, portable zip, updater signatures and latest.json — are attached to the GitHub Release. Note: the portable build does not auto-update; use an installer if you want the built-in updater.)

---

🛠️ Building from Source

Prerequisites

• Node.js (v18+)
• Rust (1.75+)
• Windows 10 / 11 Build Environment

Installation Steps

1. Clone the repository:

   git clone https://github.com/jaimitus/SpotAI.git
   cd SpotAI

2. Install frontend dependencies:

   npm install

3. Run in development mode:

   npx tauri dev

4. Build production binaries:

   npx tauri build

Publishing updates (auto-updater)

1. Generate the signing keypair once (keep the private key secret; the .pub file goes into tauri.conf.json → plugins.updater.pubkey):

   npx tauri signer generate -w ~/.tauri/spotai.key

2. Build and sign the update bundles (produces the .sig signatures next to the MSI/NSIS installers):

   export TAURI_SIGNING_PRIVATE_KEY="$(cat ~/.tauri/spotai.key)"
   export TAURI_SIGNING_PRIVATE_KEY_PASSWORD=""
   npx tauri build

3. Publish a new GitHub Release — just push the tag and the release.yml workflow does everything automatically (build, signing, automatic publishing, changelog update):

   git tag v1.4.4 && git push origin v1.4.4

   The app checks https://github.com/jaimitus/SpotAI/releases/latest/download/latest.json on startup, so that file must exist as a release asset. Two ways to get it:
   • Recommended: use tauri-action in the workflow — it builds, uploads the installers + .sig files and generates latest.json automatically.
   • Manual: upload the installers and their .sig files from src-tauri/target/release/bundle/{msi,nsis}/, then create latest.json with the static format and upload it too:

     {
       "version": "1.4.4",
       "notes": "Release notes",
       "pub_date": "2026-08-11T09:00:00Z",
       "platforms": {
         "windows-x86_64": {
           "signature": "<content of the .sig file>",
           "url": "https://github.com/jaimitus/SpotAI/releases/download/v1.4.4/SpotAI_1.4.4_x64-setup.exe"
         }
       }
     }

---

☕ Support the Project

If you find SpotAI helpful, consider supporting its development:

---

🏗️ Architecture

SpotAI
├── src-tauri/               # Native Rust Backend (Tauri v2)
│   ├── src/
│   │   ├── commands.rs      # Tauri API commands (Streaming, Ollama, Browser Launcher)
│   │   ├── native_input.rs  # Windows native input injection & paste (Ctrl+V)
│   │   ├── rag.rs           # Local RAG: sqlite-vec index + Ollama embeddings
│   │   ├── secure_store.rs  # Local DPAPI secure credential storage
│   │   ├── voice.rs         # Voice capture (cpal) + Windows SAPI dictation
│   │   ├── whisper.rs       # whisper.cpp download & offline transcription
│   │   ├── ai/              # LLM providers & streaming (OpenAI-compatible, Ollama)
│   │   └── lib.rs           # Window management, Tray icon & Global Hotkey
│   └── tauri.conf.json      # Tauri app configuration & capabilities
└── src/                     # React 19 Frontend
    ├── components/          # SpotlightWindow, ResponsePanel, SettingsModal, ActionChips
    ├── hooks/               # useLLMStream, useClipboardContext
    └── lib/                 # i18n, Prompts, Tauri API bridges, Providers

---

📄 License

This project is licensed under the MIT License.

---

Crafted with ❤️ by jaimitus