FlowOCR – Flow Launcher Plugin

Capture any region of your screen, extract the text with AI-powered OCR, and paste it anywhere — all without leaving your keyboard.

---

How it works

1. Type ocr in Flow Launcher and press Enter
2. The Windows Snipping Tool opens — select the area you want to read
3. OCR runs in the background; a notification appears when done
4. The recognised Markdown text is already in your clipboard — just paste

The OCR runs in a fully detached process so Flow Launcher stays responsive during capture and inference.

---

Backends

Backend	When to use
Ollama (default)	Local, private, no API key needed. Requires the glm-ocr model.
HuggingFace	Cloud inference via the HF serverless router. Requires a free API key.

---

Requirements

• Windows 10 / 11
• Flow Launcher ≥ 1.14
• Python 3.10+ (managed by Flow Launcher)

Ollama backend (default)

1. Install Ollama
2. Pull the OCR model:

   ollama pull glm-ocr
   

3. Make sure Ollama is running before you trigger OCR.

HuggingFace backend

Create a free account at huggingface.co, then generate a token at Settings → Access Tokens (read access is sufficient).

---

Installation

1. Download or clone this repository into your Flow Launcher plugins folder:

   %APPDATA%\FlowLauncher\Plugins\FlowOCR-1.0.0\
   

2. Restart Flow Launcher (or reload plugins via fl settings).
3. Dependencies are installed automatically on first run.

---

Settings

Open Flow Launcher settings, find Screen OCR, and configure:

Setting	Description	Default
Backend OCR	ollama or huggingface	ollama
Ollama Entrypoint	Base URL of your local Ollama server	http://localhost:11434
Hugging Face API Key	Your hf_... token (only required for the HF backend)	(empty)

You can also set the HF key via the HF_TOKEN environment variable instead of storing it in the plugin settings.

---

Troubleshooting

"Cannot connect to Ollama" Start the Ollama service (ollama serve) or check that the Entrypoint URL in settings matches your setup.

"HuggingFace authentication failed (401)" Your token is missing or expired. Regenerate it at huggingface.co and update the plugin settings.

No text detected The selected area may be too small, low-contrast, or contain no readable text. Try a larger capture region.

Log files Each run writes a timestamped log to %TEMP%\YYYYMMDD_HHMMSS.log. Check the most recent file for detailed error traces.

---

Project structure

FlowOCR-1.0.0/
├── main.py                # Plugin logic
├── plugin.json            # Flow Launcher plugin manifest
├── requirements.txt       # Python dependencies
├── SettingsTemplate.yaml  # Settings UI definition
└── Images/
    └── app.png            # Plugin icon

---

License

GPL (General Public License)