LM Studio

LM Studio is a desktop app that serves local models. Start its local server, load a model, then point EdgeQuake at it. This page lists the variables EdgeQuake reads and how to check the link.

When to use

You already use LM Studio to manage and load models.
You want a local chat and embedding server with a graphical interface.

Configure

export EDGEQUAKE_LLM_PROVIDER=lmstudio
export LMSTUDIO_HOST=http://127.0.0.1:1234
export LMSTUDIO_MODEL=your-loaded-model

Variable	Purpose	Default
`LMSTUDIO_HOST`	Server URL	`http://localhost:1234`
`LMSTUDIO_MODEL`	Chat model you loaded. LM Studio serves only what is loaded.	`gemma2-9b-it`
`LMSTUDIO_EMBEDDING_MODEL`	Embedding model, used when `EDGEQUAKE_EMBEDDING_PROVIDER=lmstudio`	`nomic-embed-text-v1.5`
`LMSTUDIO_EMBEDDING_DIM`	Embedding vector size	`768`

LM_STUDIO_BASE_URL is not read. Use LMSTUDIO_HOST.

Models

Chat: load the model in LM Studio, then set LMSTUDIO_MODEL to its name.
Embeddings: load an embedding model in LM Studio, then set EDGEQUAKE_EMBEDDING_PROVIDER=lmstudio and LMSTUDIO_EMBEDDING_MODEL.

Verify

curl http://127.0.0.1:1234/v1/models
curl http://127.0.0.1:8080/health

/health reports degraded when LM Studio does not answer on /v1/models. The probe (POST /api/v1/providers/test with "shape":"openai_chat") lists the same models. See Test a server.

%%{init: {"theme":"base","themeVariables":{"primaryColor":"#E0E7FF","primaryBorderColor":"#6366F1","primaryTextColor":"#1E1B4B","secondaryColor":"#D1FAE5","secondaryBorderColor":"#10B981","secondaryTextColor":"#064E3B","tertiaryColor":"#FEF3C7","tertiaryBorderColor":"#F59E0B","tertiaryTextColor":"#6B7A90","lineColor":"#7A889C","clusterBkg":"rgba(99,102,241,0.07)","clusterBorder":"#A5B4FC","noteBkgColor":"#FEF9C3","noteTextColor":"#422006","textColor":"#6B7A90","titleColor":"#6B7A90","signalColor":"#7A889C","signalTextColor":"#6B7A90","loopTextColor":"#6B7A90","edgeLabelBackground":"#F1F5F9","actorLineColor":"#94A3B8"}}}%%
%% eq-theme:v1
flowchart LR
  A["EdgeQuake probe"] --> B["LM Studio GET /v1/models"]
  B --> C{"LMSTUDIO_MODEL listed?"}
  C -- "Yes" --> D["Chat check runs"]
  C -- "No" --> E["Probe reports model_not_found"]
%% eq-classes
classDef eqLlm fill:#FEF3C7,stroke:#F59E0B,color:#451A03
class B eqLlm

In Docker, use http://host.docker.internal:1234. quickstart.sh --provider lmstudio --base-url URL sets LMSTUDIO_HOST for you.

Troubleshoot

Probe reports model_not_found. The name in LMSTUDIO_MODEL is not one LM Studio lists. Copy the name from /v1/models.
/health is degraded. LM Studio’s local server is stopped, or it listens on another port. Start the server in LM Studio and match the port.
Embeddings fail. No embedding model is loaded in LM Studio, or EDGEQUAKE_EMBEDDING_PROVIDER is not set to lmstudio.