Skip to content
Yttri Docs

LM Studio

LM Studio runs models on your own machine behind a local OpenAI-compatible server. Yttri connects to it like any other provider.

Connecting

Section titled “Connecting”
  1. In LM Studio, load a model and on the Developer tab turn on Start Server — by default it listens on http://localhost:1234.
  2. In Yttri: Settings → Assistant → Models → Add provider → pick the LM Studio preset (the Base URL and API type fill in on their own) → “Test connection”.
  3. No API key needed — the server is local.

The “Local provider” flag is set automatically from the localhost address — it controls whether personal data is masked before sending (see Built-in models and Models, router and privacy).

Known quirks

Section titled “Known quirks”

Some LM Studio versions reject a system message that isn’t first in the conversation and require a tool choice to be forced for tool calling — Yttri already works around both on its side. If tool calls still look odd, that’s more likely an LM Studio/model quirk than a Yttri bug.

If it won’t connect

Section titled “If it won’t connect”
  • Make sure the server is actually running in LM Studio (Developer tab, green indicator) — it doesn’t start automatically with the app.
  • A model needs to be loaded into memory — “Test connection” fails without one.
  • The default port is 1234; if you changed it in LM Studio’s settings, update the Base URL in Yttri to match.