AI Model Manager

A centralized desktop application for managing local AI models across different tools and frameworks.

The Problem

Anyone using local AI models knows the pain: models scattered everywhere. Ollama stores weights in one place, ComfyUI in another, LM Studio has its own cache, Kdenlive downloads Whisper to yet another location, and HuggingFace has its own cache directory. You end up with:

• Duplicated models consuming disk space (the same Whisper model downloaded 3 times)
• No unified view of what you actually have installed
• Confusion when connecting tools to models ("where did I put that?")
• Difficulty cleaning up old or unused models

The Solution

AI Model Manager provides a single source of truth for all your local AI models:

• One base directory for all models, organized by category
• A GUI to browse, search, and manage your model library
• Support for pointing any compatible tool to your centralized store
• Metadata tracking via SQLite database
• Hugging Face integration for downloads

Features

Core Features

• Centralized Storage: Define a base path (e.g., ~/AI/models) as your single model store
• Category-Based Organization: Models organized by type (LLMs, diffusion, STT, TTS, etc.)
• GUI Interface: Desktop application for browsing and managing models
• SQLite Database: Track model metadata, locations, and tags
• Hugging Face Integration: Download models directly with HF token support

Model Categories

Category	Description	Examples
llm/	Large Language Models	Llama, Mistral, Qwen
llm/gguf/	Quantized LLMs	GGUF format models
llm/finetunes/	Fine-tuned LLMs	Custom trained models
embedding/	Embedding Models	nomic-embed, bge
diffusion/	Image Generation	Stable Diffusion, FLUX
diffusion/lora/	LoRA Adapters	Style/character LoRAs
diffusion/controlnet/	ControlNet Models	Pose, depth, canny
diffusion/vae/	VAE Models	SD VAEs
stt/	Speech-to-Text	Whisper, Whisper fine-tunes
tts/	Text-to-Speech	Fish Speech, Kokoro
multimodal/	Vision-Language Models	LLaVA, Qwen-VL
misc/	Other Models	Punctuation restoration, etc.

Planned Features

• Duplicate Detection: Find and remove duplicate models
• AI Assistant: Chat with a local agent to help clean up and organize
• Import Wizard: Scan existing tool caches and consolidate
• Export/Sync: Backup model library metadata

Installation

Coming soon

Usage

Setting Up Your Model Store

1. Launch AI Model Manager
2. Configure your base path (e.g., ~/AI/models)
3. The app creates the category folder structure
4. Add your Hugging Face token for authenticated downloads

Pointing Tools to Your Store

Most AI tools support custom model paths:

Ollama:

export OLLAMA_MODELS=~/AI/models/llm

ComfyUI:

# extra_model_paths.yaml
base_path: ~/AI/models/diffusion

LM Studio: Configure model directory in settings to ~/AI/models/llm/gguf

Tech Stack

To be determined - Candidates:

• GUI: Electron, Tauri, PyQt, or web-based (Flask/FastAPI + browser)
• Database: SQLite
• Language: Python or TypeScript

Contributing

Contributions welcome! This project is open source.

License

TBD

Acknowledgments

Inspired by the organizational approach of ComfyUI's model directory structure and tools like OllaMan and civitai-data-manager.