Skip to content

Home

Jump to bottom
martymcfly1 edited this page Feb 28, 2026 · 1 revision

ImageFlow

Extract structured data from images and PDFs using local Vision Language Models.

What is ImageFlow?

ImageFlow is a macOS application that leverages the power of local Vision Language Models (VLMs) to automatically extract structured data from images and single-page PDFs. Whether you need to pull data from scanned documents, screenshots, receipts, or forms, ImageFlow makes it simple—no cloud services or API keys required.

Features

Local AI Processing

  • Works entirely offline with your own AI models
  • Supports LM Studio and Ollama as providers
  • Your data never leaves your machine

Document Support

  • Images: JPG, PNG, WebP, GIF
  • Single-page PDFs
  • Drag-and-drop or file picker upload

Flexible Data Extraction

  • Define custom columns with descriptions
  • Save extraction templates for reuse
  • Process single or batch documents

Export Options

  • Export to CSV or Excel format
  • Review and edit extracted data before exporting

Getting Started

1. Install Prerequisites

Download and install one of these local AI providers:

  • LM Studio - User-friendly interface for running LLMs
  • Ollama - Lightweight AI runtime

Load a vision-capable model such as:

  • LLaVA
  • MiniCPM-V
  • Qwen2-VL
  • or any other VLM compatible with Ollama or LM Studio

2. Configure ImageFlow

  1. Open ImageFlow
  2. In Settings, select your provider (LM Studio or Ollama)
  3. Enter the provider URL:
    • LM Studio: http://localhost:1234/v1
    • Ollama: http://localhost:11434
  4. Click "Test Connection" to verify
  5. Select your model from the dropdown

3. Extract Data

  1. Add Documents - Drag and drop images or PDFs
  2. Design Your Table - Define columns for the data you want to extract
  3. Process - Click Start Processing and watch the magic happen
  4. Export - Save your extracted data as CSV or Excel

System Requirements

  • macOS (Apple Silicon or Intel)
  • LM Studio or Ollama running locally
  • A vision-capable model loaded in your provider

Privacy

All processing happens locally on your machine. No data is sent to external servers, making ImageFlow ideal for sensitive documents.

Tech Stack

  • Electron - Desktop application framework
  • React 19 - User interface
  • Tailwind CSS - Styling
  • Zustand - State management
  • Sharp - Image processing
  • pdf.js - PDF rendering

License

MIT License - See LICENSE file for details.

Support

For issues, feature requests, or questions, please open an issue on GitHub.

ImageFlow - Extract data from images with local AI

Clone this wiki locally