Download and manage the local language model that the AI features on a SynapseOS desktop use. It fetches, verifies and removes them, and nothing more.
syn-model list # what is on offer
syn-model status # what is installed, and where
syn-model download # the default, mistral-7b
syn-model download tiny -y # a specific one, without prompting
syn-model remove| name | what it is | size |
|---|---|---|
mistral-7b |
Mistral 7B Instruct Q4_K_M — the recommended one | ~4.1 GB |
phi3 |
Phi-3 Mini 4K Instruct Q4 — noticeably weaker | ~2.2 GB |
tiny |
Qwen2 0.5B Q4 — fits anywhere, and answers like it | ~400 MB |
A smaller model is not just a smaller download. Everything downstream — a natural-language shell, a coding assistant, a desktop AI panel — follows instructions worse with one, and the failure looks like the tool being broken rather than the model being small.
syn-model fetch TOKEN runs a queued download as root, driven by
syn-model-download@TOKEN.service and authorised through a polkit rule. That
is how a desktop's downloader asks for a model without any part of the UI
running privileged.
Models are large and the download is resumable; status is the way to see
what actually landed rather than assuming a finished command means a finished
file.
git clone https://github.com/velle999/syn-model
cd syn-model && makepkg -simakepkg fetches the source for this PKGBUILD's exact version from this
repository's releases, so a clone can only ever build the source it was
written against. .SRCINFO lists what it needs.
Developed in the SynapseOS monorepo,
in syn-model/. This repository is generated from it — the PKGBUILD, a
generated .SRCINFO and this README — so issues and patches belong there.
syn-model 0.1.0-6 · GPL-2.0-or-later