EZmodL

GGUF Model Manager for llama-swap — browse HuggingFace, download, manage models, all in one place.

EZmodL

A lightweight web UI to manage your local GGUF model collection.


✨ Features

📦 Compatibility

Tested on:

Should work, but untested — same apt base, no reports either way:

Non-apt distros (Fedora, Arch, openSUSE): system dependencies must be installed manually.

🚀 Quick install

git clone https://github.com/miradorventus/EZmodL.git
cd EZmodL
EZMODL_LOCAL_INSTALL=1 bash install.sh

Or as a one-liner:

curl -fsSL https://raw.githubusercontent.com/miradorventus/EZmodL/main/install.sh | bash

📌 The installer downloads the files, installs system dependencies (via pkexec), sets up the menu and desktop entries, and tells you if anything is missing.

🔧 Requirements

EZmodL is designed to work with llama-swap, so you need:

For a fully automated llama.cpp + llama-swap setup on AMD, see ollama-amd-plug-and-play.

🎬 Usage

  1. Launch EZmodL from the Applications menu
  2. Search for a model (e.g. Qwen3) or paste a repo (bartowski/Qwen_Qwen3-14B-GGUF)
  3. Set your KV cache VRAM target
  4. Tick the quantization(s) you want
  5. EZmodL handles the download, detects duplicates, adds the entry to the YAML, and reloads llama-swap

🛠️ Configuration

All configuration lives in ~/.ezmodl/:

~/.ezmodl/
├── config/
│   ├── preferences.json    # VRAM target, sort order, curators_only, theme
│   └── curators.txt        # Editable curator list
├── queue/                  # Download queue and history
└── logs/                   # Activity log

Editing the curator list (add or remove):

nano ~/.ezmodl/config/curators.txt

🎨 Theming

The CSS is exposed at ~/.ezmodl/static/style.css. CSS variables sit at the top of the file:

:root[data-theme="dark-futurist"] {
  --bg-base: #0a0e1a;
  --text-accent: #00d9ff;
  /* ... */
}

You can build your own theme by adding a :root[data-theme="my-theme"] block.

🗂️ Architecture

Component Role
server.py Flask backend, exposes a REST API
lib/hf_api.py HuggingFace API wrapper (search + tree)
lib/vram_classify.py AMD/NVIDIA VRAM detection and quantization classification
lib/disk_search.py Multi-folder search and symlink creation
lib/yaml_helper.py Add/remove models in llama-swap.yaml
lib/parse_quant.py Extracts the quantization from a filename
templates/index.html Main UI (collapsible sections)
static/app.js Vanilla JS, fetches the REST endpoints
ezmodl.sh Launcher, OllamaUI pattern (lockfile, integrity check)

❌ Uninstall

cd EZmodL
bash uninstall.sh

GGUF models in ~/llm-models/ are kept. llama-swap.yaml is left untouched — clean it up yourself if needed.

📋 V1 limitations

🤝 Contributing

Issues and PRs welcome. This project is deliberately simple and minimal — let’s keep it that way.

📜 License

MIT — see LICENSE.

🙏 Credits


EZmodL — because managing GGUFs shouldn’t be a chore.