Skip to content

Settings: add LM Studio, Ollama and Atomic Chat as local apps, never as cloud - #1169

Merged
sosidudku1 merged 3 commits into
rel/xp-buildfrom
bugs-1401-lmstudio-local-app
Nov 12, 2026
Merged

sosidudku1 merged 3 commits into
rel/xp-buildfrom
bugs-1401-lmstudio-local-app

Conversation

@sosidudku1

@sosidudku1 sosidudku1 commented Nov 10, 2026 •

Copy link
Copy Markdown
Collaborator

A Windows user on Desktop 0.0.8 typed LM Studio's address (http://127.0.0.1:1234) into Settings › Custom server and got "OpenAI-compatible server — add it as a cloud provider?". They wanted LM Studio as a local worker for Fusion, so the word "cloud" stopped them (ATO-1401).

What changes

  • The Custom server prompt recognises LM Studio by its port, as it already did for Ollama and Atomic Chat. All three are matched only on a loopback address: a server on another machine is a plain provider. For any of the three it says "This is LM Studio. Add it as a local app?" and the button is Add LM Studio. One click adds it at the address typed and goes straight to its model list, since a local app needs no key. An unknown OpenAI-compatible server gets "Add it as a provider?". The prompt no longer says "cloud".
  • Settings › Custom server shows "Use an app you already run:" with Add Ollama / Add LM Studio / Add Atomic Chat for the ones not added yet. Settings › Local models gets Add LM Studio and Add Atomic Chat next to the existing Add Ollama.
  • In the Fusion seat picker, a provider on this computer (those three, or any loopback server) reads "An app on this computer" instead of "Cloud provider". This is judged by the provider's address, not its preset id.

Checks added

  • test/settings-local-apps.test.mjs: ports to presets (and none for another host), no "cloud provider" in the prompt, the Fusion seat rows built in a vm.

Not done here

  • Added apps still show in the "Cloud providers" list of Settings, since that is where provider entries live. Moving them is a bigger change; the first-run part of this is in the onboarding PR (ATO-816 slices).

Linear: ATO-1401

Checked on the second Mac on 12.11.2026 in the combined run with the other open PRs: build, lint, test types, vitest, desktop unit tests, desktop smoke and live smoke all green.

…as cloud

A Windows user typed LM Studio's address (127.0.0.1:1234) into Settings ›
Custom server and was asked to "add it as a cloud provider". They wanted it
as a local worker for Fusion and stopped there.

- The Custom server prompt names the app (LM Studio is now recognised by
  its port too) and asks "Add it as a local app?". One click adds it at the
  address typed and goes straight to its model list: a local app needs no
  key. No word "cloud" in the prompt.
- Custom server and Local models show "Add LM Studio" / "Add Atomic Chat"
  (and Ollama on Custom server) while they are not added yet.
- In the Fusion seat picker a provider on this computer reads "An app on
  this computer", not "Cloud provider".

ATO-1401
…y address

Ollama's port matched on any host, so a LAN box at :11434 was added as an
app on this computer, keyless, and shown as local in Fusion. All three
presets now need a loopback address. The Fusion seat label judges a
provider by its address, not its preset id. The test builds the seat rows
in a vm instead of matching source text.
@sosidudku1
sosidudku1 merged commit 92f7acd into rel/xp-build Nov 12, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant