VisionTagger

Note: This project is not yet available on GitHub.

Overview

VisionTagger is a desktop application that uses Ollama's vision models to analyze images and generate descriptive tags and filenames. Point it at a folder of images and it will examine each one, suggest tags, and offer to rename files based on their content—all running locally without cloud dependencies.

Key Features

• Vision AI Analysis: Uses Ollama vision models to understand image content.
• Automatic Tagging: Generates descriptive tags for each image.
• Smart Renaming: Suggests filenames based on image content.
• Batch Processing: Handle entire folders of images at once.
• Local Processing: All AI runs locally via Ollama—no data leaves your machine.
• Flet UI: Cross-platform desktop interface.

Technical Architecture

VisionTagger connects to a local Ollama instance running vision-capable models. It processes images through the vision API, extracts descriptions, and applies naming conventions based on the results.

Core components:

• Ollama Client: Connects to local Ollama for vision model access.
• Image Processor: Handles image loading and format conversion.
• Tag Generator: Parses vision model output into structured tags.
• Rename Engine: Applies naming rules based on generated tags.

Technology Stack

• UI Framework: Flet for cross-platform desktop
• AI Backend: Ollama with vision models (LLaVA, etc.)
• Language: Python
• Image Handling: Pillow for format support

Current Status

Prototype with core tagging working. Note: renaming is currently destructive without undo—use with caution on copies. Planning to add rename preview and undo support.