sammcj/gollama

Gollama – a terminal UI for managing Ollama models

What it does – Gollama is a small Go program that talks to the local Ollama API and lets you browse, start, stop, edit, copy, delete and push models from a terminal‑based UI. It also offers a handful of command‑line flags for quick one‑off actions (list, search, edit, unload, vRAM estimation, …).

Why it matters – Ollama is a popular way to run LLMs locally. Managing dozens of models (different sizes, quantisations, families) can become cumbersome. Gollama gives you a searchable, sortable list, shows metadata (size, quantisation, family, last‑modified), and can launch ollama run or unload models with a single keystroke. The built‑in vRAM estimator helps you decide whether a model will fit on your GPU/CPU before you pull it.

Key features

• TUI with hot‑keys for selecting, sorting, filtering, editing Modelfiles, copying/renaming, deleting, running and unloading models.
• Command‑line shortcuts (gollama -l, -s, -e, --vram …) for scripting.
• vRAM usage calculator for Ollama and Hugging‑Face models, including context‑length and quantisation trade‑offs.
• Configurable JSON file (~/.config/gollama/config.json) for default sort, visible columns, editor, theme, API endpoint, etc.
• Light‑weight theming support (dark, light‑neon, custom JSON themes).
• Logging to ~/.config/gollama/gollama.log with adjustable log levels.

How to get it – Install with Go (go install github.com/sammcj/gollama/v2@latest), via the provided curl script, or download a pre‑built binary from the releases page. The repo also contains a make build recipe for building from source.

Who might use it – Anyone running Ollama on macOS or Linux who prefers a keyboard‑driven interface over the default CLI, especially power users who juggle many models and need quick insight into memory requirements.

---

Short‑answer summary – Gollama is a genuine, open‑source utility for managing locally‑served LLMs via Ollama. It does not train or serve models itself, but it streamlines model administration, inspection, and resource‑planning for AI practitioners.