Miaw AI secretary

Product Introduction

1. Definition: Miaw AI secretary is an open-source, Electron-based desktop application designed as a lightweight, floating artificial intelligence interface. It functions as a cross-platform (Windows-first) AI assistant that leverages the React framework for its user interface and integrates with various Large Language Model (LLM) providers, specifically optimized for local execution via Ollama and LiteLLM-compatible endpoints.

2. Core Value Proposition: The project exists to provide a non-intrusive, "always-available" AI sidekick that eliminates the friction of switching between browser tabs or heavy applications. By utilizing a global shortcut (Ctrl+Shift+Space) and a floating UI, Miaw AI secretary enables users to access high-performance AI models—including local, privacy-focused LLMs—directly from their desktop environment, enhancing productivity through immediate contextual assistance and system-level integration.

Main Features

1. Electron-Based Desktop Shell with Tray Integration: Miaw AI secretary uses the Electron framework to wrap a modern React-based UI into a native Windows application. It features a system tray icon for background operation and a global keyboard listener (Ctrl+Shift+Space) that triggers a floating overlay. This architecture ensures that the AI is accessible instantly without cluttering the taskbar, providing a seamless user experience similar to Spotlight or Raycast.

2. Local Model Support via Ollama and LiteLLM: The application is technically architected to support local inference, reducing latency and increasing data privacy. It includes built-in command surfaces for Ollama (ask_ollama) and supports LiteLLM-compatible OpenAI endpoints. This allows users to run models like Gemma-3, Llama 3, or Mistral locally on their hardware, governed by configurations set within a .env environment file.

3. Advanced Command Surface and Tauri-Compatible Shims: Although built on Electron, the application utilizes specialized shims for @tauri-apps/api core functions (event, window, and dpi). This technical bridge allows it to reuse logic from Tauri-based projects while maintaining Windows-first stability. The command surface includes robust features such as chat streaming, conversation persistence (save, list, load, and delete conversations), and automated title generation for chat history management.

4. Integrated Screen Intelligence (Screenshot Capture): Miaw AI secretary features native commands for screenshot interaction (capture_screenshot_command and capture_full_screen_command). This allows the assistant to potentially process visual information from the user's screen, facilitating "visual-to-text" workflows where the AI can analyze active windows, code snippets, or UI designs in real-time.

Problems Solved

1. Context Switching Fatigue: Developers and researchers often lose focus when switching between their IDE/workspace and a web-based AI chat. Miaw AI solves this by providing a floating overlay that appears directly over the active work environment.

2. Data Privacy and API Costs: By supporting local LLM integration through Ollama, the application addresses the privacy concerns of enterprise users and the recurring costs associated with cloud-based AI subscriptions. Sensitive data remains on the local machine.

3. Target Audience:

• Software Developers: Requiring quick code explanations or debugging assistance without leaving the editor.
• Privacy Enthusiasts: Users who prefer running open-source models (Gemma, Llama) locally.
• Power Users: Individuals who utilize global hotkeys to streamline their digital workflows.
• Technical Writers: Needing a quick interface for drafting or summarizing content based on on-screen information.

4. Use Cases:

• Instant Code Refactoring: Highlighting a block of code and summoning the assistant via hotkey to suggest optimizations.
• Local Knowledge Base: Interacting with a locally hosted LLM to query private documents without uploading them to the cloud.
• Visual Debugging: Using the screenshot capture feature to send a UI bug report to the AI for immediate feedback or CSS suggestions.

Unique Advantages

1. Differentiation: Unlike web-based wrappers, Miaw AI secretary is a native-feeling desktop utility that recreates the premium experience of macOS-specific AI tools (like Thuki) for the Windows ecosystem. It provides a "Windows-first" experience with specialized installer (.exe) and portable build options.

2. Key Innovation: The project’s unique innovation lies in its hybrid architecture—reusing a React UI and a Tauri-compatible command surface within an Electron shell. This approach allows for rapid development and compatibility with existing AI UI components while leveraging Electron's mature ecosystem for Windows window management and global input handling.

Frequently Asked Questions (FAQ)

1. How do I connect Miaw AI secretary to a local model? To use local models, you must have Ollama or a LiteLLM server running. You configure the connection by copying the .env.example file to .env and setting the THUKI_API_BASE_URL to your local endpoint (e.g., http://127.0.0.1:4000) and specifying your preferred model in THUKI_SUPPORTED_AI_MODELS.

2. Does Miaw AI secretary support GPT-4 or other cloud providers? Yes, because the backend is compatible with OpenAI-style endpoints via LiteLLM, you can connect it to cloud providers by entering the appropriate API key and base URL in the environment configuration, allowing for a mix of local and cloud-based intelligence.

3. Is Miaw AI secretary compatible with macOS or Linux? While the current development focus is "Windows-first" and the README highlights Windows build scripts (Portable EXE and Installer), the underlying technologies (Electron, React, Vite) are inherently cross-platform. However, certain window-handling gestures and shell commands are currently optimized for the Windows environment.