cloudparts

Alpaca

A GNOME desktop app for Linux that chats with local models through Ollama, which it can install and run for you, and with cloud models through OpenAI-compatible APIs using your own keys.

What it is
Running models locally
Type
Open source
Made by
Jeffry Samuel Eduarte Rojas (Jeffser)
License
GPL 3.0 or later
Facts checked
13 October 2026
Website Source code

What it is

Alpaca is a native GNOME app, written in Python with GTK and libadwaita and distributed officially as a Flatpak on Flathub. Text generation goes through an instance, and one is active at a time. Ollama (Managed) is the recommended one: Alpaca starts Ollama only while it needs it and lets you pick the model folder and GPU overrides. Ollama (External) connects to a server you already run, for example Ollama in a Docker container; Ollama (Cloud) uses Ollama’s hosted service with an API key; and 19 named OpenAI-compatible providers, among them OpenAI, Google Gemini, DeepSeek and Groq, plus a generic OpenAI-compatible instance, work with your own API key. Since Alpaca 9 (January 2026), a built-in Ollama Manager installs Ollama inside the Flatpak sandbox, adds ROCm for AMD GPUs and checks for Ollama updates, replacing the earlier Ollama and ROCm Flatpak extensions.

Chats are kept in SQLite files on the device, and with a local model nothing leaves the computer once the model is downloaded. Several models can answer in the same conversation, and Ollama instances can create custom models from a GGUF file or an existing model. Attachments include images for vision models, plain text and PDF documents, the text of a web page and YouTube transcripts. Tools let a model run a web search, a terminal command or a background remover; activities are mini-apps in a sidebar or window, such as a terminal (local or over SSH), a code runner, a web browser and Live Chat for voice conversations. Speech recognition (OpenAI Whisper) and text to speech (Kokoro) also run locally; the Flatpak bundles both and downloads their models on demand. On GNOME, a search provider starts a Quick Ask from the shell’s search bar.

Why teams pick it

  • A native desktop app with no web server to run: one Flatpak, and it can install, start and update Ollama itself
  • Local by default: chats in SQLite on the device and full offline use once models are downloaded
  • Mixes local and cloud models: Ollama plus OpenAI-compatible providers with your own keys, and several models in one chat
  • Free software under GPL 3.0 or later, with frequent releases (9.2.5 in August 2026) and community translations into more than 20 languages

Watch out for

  • Linux first: Flathub is the only official channel, AUR, Nix and Snap packages are community-maintained, and the macOS build announced with 3.0 was experimental
  • Ollama is the best-supported backend; the docs list most OpenAI-compatible providers as untested and Anthropic, OpenRouter, Mistral and Fireworks as inconsistent
  • Inside the Flatpak sandbox the docs recommend the terminal activity’s SSH mode for running commands on the host
  • Speed and model size depend on your own hardware, and AMD GPUs need ROCm, installed separately through the Ollama Manager

Alternatives

What all running models locally have in common →Compare side by side →
LM StudioCommercialA free, closed-source desktop app and headless daemon that downloads open-weight models from Hugging Face, runs them on your own machine with llama.cpp or Apple’s MLX, and serves them on OpenAI- and Anthropic-compatible endpoints at localhost:1234.Open WebUICommercialA self-hosted web app for chatting with models served by Ollama or any OpenAI-compatible API, with users and groups, document retrieval, tools and web search, under a source-available license that requires keeping its branding above 50 users.

How it fits with other technologies

Install from Flathub · bash

flatpak install com.jeffser.Alpaca
flatpak run com.jeffser.Alpaca

# then: create an Ollama (Managed) instance and let Alpaca install Ollama,
# or add an Ollama (External) or OpenAI-compatible instance with your API key

See it in a real stack

No architecture uses Alpaca yet. Drop it into the playground and wire it up yourself.Open playground →
Where it came from2024–2026

Jeffry Samuel Eduarte Rojas, who publishes as Jeffser from Costa Rica, released Alpaca in May 2024 as an Ollama client for GNOME: a beginner-friendly way to chat with local models, open source and powered by Ollama. Version 1.0.0 in July 2024 was announced as the stable release, with almost 15,000 users on Flathub.

Major versions followed every few months. 2.0 (September 2024) rewrote the model, message and chat systems and added experimental AMD GPU support; 3.0 (December 2024) brought a stable Snap and an experimental macOS build; 5.0 (February 2025) added an instance manager and the first cloud providers, OpenAI, Gemini, Together AI and Venice; 6.0 (April 2025) added tools, speech recognition and text to speech; and 8.0 (September 2025) added folders and activities such as a terminal, code runner and web browser.

Alpaca 9 (January 2026) replaced the Ollama and ROCm Flatpak extensions with a built-in Ollama Manager, so users get new Ollama releases without waiting for the extension to be updated. The project now describes itself as a “local and online AI hub.”

Why it’s built this way

Build on Ollama rather than a runtime of its own
Ollama already pulls and runs models, so Alpaca could focus on an easy interface and on managing Ollama’s process so it runs only when needed.
Flathub as the official channel
One Flatpak covers the distributions Flathub supports; Alpaca welcomes repackaging but doesn’t vouch for community packages.
Cloud providers through the OpenAI library
Every non-Ollama instance uses the OpenAI Python library, so one code path covers many providers; the cost is testing, since checking each one needs paid tokens.
An Ollama Manager instead of Flatpak extensions
Users can install the newest Ollama the moment it comes out instead of waiting for a new extension release.

Timeline

  1. 2024

    First release, 0.1.1 (13 May)

  2. 2024

    1.0.0, the first stable release (July)

  3. 2025

    5.0 adds an instance manager and cloud providers (February)

  4. 2025

    8.0 adds folders and activities (September)

  5. 2026

    9.0 replaces the Flatpak extensions with the Ollama Manager (January)

  6. 2026

    9.2.5 released (August)

Sources10 · facts checked 13 October 2026

Improve this page on GitHub