How ryuumeow/neverthinkautoreply works
Project type and purpose
• This is a Windows-only Python desktop application (“Never Think Auto Reply”) that listens for a global hotkey, captures text or images, runs AI/OCR processing, and automatically pastes reply content.
File structure
• Top-level scripts:
• NeverThinkAutoReply.py (entry point, GUI + orchestration)
• ocr_download.py (download OCR models)
• config.ini (runtime settings)
• src/ – modular code:
• src/api/llm.py – AI service adapter (OpenAI, Gemini, DeepSeek)
• src/configs.py – configuration loader
• src/utils/ – logging, file naming, clipboard copy helpers
Readme insights
• GUI framework: PySide6
• System-level hooks: pynput (hotkey), pyperclip (clipboard), pyautogui (paste/click), windows-toasts (notifications), pywin32
• OCR: easyocr + OpenCV
• AI backends: OpenAI API, HuggingFace model loading support
Main components and interactions
┌──────────────────┐ ┌───────────────────────────┐
│ 1. Hotkey Listener │───► │ 2. UI Dialog & Mode Selector │
└──────────────────┘ └───────────────────────────┘
│ │
│ ▼
│ ┌────────────────────┐
│ │ 3. Core Orchestrator │
│ └────────────────────┘
│ │
│ ┌──────────┬─────────┴─────────┬───────────┐
│ ▼ ▼ ▼ ▼
┌────────────────┐ ┌──────────┐ ┌──────────────────────┐ ┌─────────────────┐
│4a. Clipboard │ │4b. OCR │ │4c. LLM API Adapter │ │4d. Asset Manager │
│ Manager │ │ Module │ │ (OpenAI/HF/DeepSeek) │ │ (prompts/mygo/…) │
└────────────────┘ └──────────┘ └──────────────────────┘ └─────────────────┘
│ │ │ │
└─────────────┴────────────────┴───────────────────────┘
▼
┌────────────────────┐
│5. Automation & UI │
│ output (pyautogui,│
│ windows-toasts) │
└────────────────────┘
│
▼
┌────────────────────┐
│6. Logging & Config │
└────────────────────┘
• Data flows from the hotkey event through the orchestrator into the selected mode (normal, rebut, sarcasm, MyGo/Mujica image, OCR).
• In OCR mode, image data routes via the OCR Module before entering the orchestrator.
• The LLM API Adapter is invoked for text-based modes; Asset Manager provides pre-bundled images for image modes.
• Output is automatically pasted and a Windows toast is shown. All steps are configurable via config.ini; operations are logged.
Architectural patterns and design principles
• Layered modular monolith: UI layer, orchestration layer, service adapters (LLM, OCR), utility layer (clipboard, logging), configuration
• Separation of concerns: each module has a single responsibility (hotkey, UI, AI, OCR, assets, notifications)
• Configuration-driven: behavior and API keys controlled by config.ini
• Extensibility points: adding new reply modes, swapping AI backends, dropping in new assets