How to Configure Local Models with LM Studio in G0DM0D3: A Complete Setup Guide

G0DM0D3 supports local inference through any OpenAI-compatible server, allowing you to route all LLM operations—including classification, judging, and refinement—to a private LM Studio instance running on localhost.

The G0DM0D3 repository by elder-plinius provides a privacy-focused interface that can operate entirely offline by connecting to local inference engines. When you configure local models with LM Studio, the application bypasses cloud providers like OpenRouter and Venice, ensuring sensitive data never leaves your machine while maintaining full access to features like TASTEMAKER judging and Liquid refinement.

Prerequisites: Launching the LM Studio Server

Before configuring G0DM0D3, start LM Studio and load your desired model. LM Studio exposes an HTTP server at http://localhost:1234/v1 by default, implementing the standard OpenAI-compatible /v1/models and /v1/chat/completions endpoints. Verify the server is running by testing http://localhost:1234/v1/models in your browser or via curl.

Configuring the Local Endpoint in G0DM0D3

Step 1: Access the Settings Modal

Open the Settings modal in the G0DM0D3 web interface. The component SettingsModal.tsx initializes state hooks for your local configuration at lines 1648–1654, reading the current localUrl and localKey values from the global store. These fields control where the application sends inference requests when operating in private mode.

Step 2: Enter the LM Studio Base URL

Locate the Local Model URL input field and enter http://localhost:1234/v1 (or your custom LM Studio URL). In src/components/SettingsModal.tsx at line 1703, this input binds directly to the localUrl state via setLocalUrl, updating the target endpoint in real time.

Step 3: Test and Discover Available Models

Click the Test & Discover Models button to verify connectivity. This triggers a request to ${baseUrl}/models handled by src/lib/openrouter.ts at line 45. The function parses the JSON response and populates the model selector with available local models, ensuring you can choose specific variants loaded in LM Studio.

Step 4: Enable Local-Only Mode

Toggle the Local-only mode switch to force all traffic to your private server. This setting updates the localOnly boolean flag defined in src/store/index.ts at line 12. When active, the request layer excludes OpenRouter and Venice from every "ULTRAPLINIAN" race, disables telemetry to remote G0DM0D3 services, and routes all internal LLM calls—including query classification, refusal checks, and coaching—to your local endpoint.

How Request Routing Works Under the Hood

The chatCompletion function in src/lib/openrouter.ts dynamically selects the appropriate base URL based on your configuration. When localOnly is true, it uses store.getState().localUrl instead of the cloud provider URL.

export async function chatCompletion(payload: ChatPayload) {
  const base = store.getState().localOnly
    ? store.getState().localUrl      // ← LM Studio endpoint
    : store.getState().ultraplinianApiUrl; // ← cloud provider

  const resp = await fetch(`${base}/chat/completions`, {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify(payload),
  });
  return resp.json();
}

This conditional routing ensures that enabling local mode affects every downstream operation, from initial query classification to final Liquid refinement passes, without requiring code changes or restarts.

Handling CORS for Hosted vs. Local Deployments

If you access G0DM0D3 from the hosted page at https://godmod3.ai, you must configure LM Studio to allow cross-origin requests from that domain. Start LM Studio with the --origin flag:

lm-studio start --origin https://godmod3.ai

Alternatively, run the UI locally to avoid CORS restrictions entirely:

python3 -m http.server 8080

Then navigate to http://localhost:8080 and point the settings to http://localhost:1234/v1. Local deployments require no additional CORS configuration since both the UI and inference server share the same origin.

Troubleshooting Common Connection Issues

Connection failed / Failed to fetch – Verify LM Studio is running and accessible at your configured URL. If using the hosted UI, confirm you started LM Studio with the appropriate --origin flag to enable CORS for https://godmod3.ai.

No models returned – Ensure a GET request to <your-base-url>/models returns a JSON object containing a data array with objects structured as { "id": "model-name" }. LM Studio must have a model actively loaded to populate this list.

Slow responses or timeouts – Reduce the Max Tokens setting in G0DM0D3 or unload unused models from LM Studio to free up VRAM. Local inference speed depends heavily on your hardware configuration and model quantization level.

Summary

• G0DM0D3 connects to LM Studio via OpenAI-compatible endpoints at /v1/models and /v1/chat/completions.
• Configure the connection in SettingsModal.tsx by setting the base URL and enabling Local-only mode stored in src/store/index.ts.
• The localOnly flag forces all inference requests to bypass cloud providers, disabling external telemetry while preserving full functionality.
• Persistence is handled automatically through localStorage serialization in src/store/index.ts at line 78.
• CORS configuration is only required when accessing the UI from https://godmod3.ai; local deployments work without additional flags.

Frequently Asked Questions

What endpoints must LM Studio expose for G0DM0D3 compatibility?

LM Studio must provide OpenAI-compatible /v1/models and /v1/chat/completions endpoints. The application queries /v1/models to populate the model selector and sends all inference requests to /v1/chat/completions with standard ChatML formatting.

Does enabling local-only mode disable cloud providers entirely?

Yes. When the localOnly flag in src/store/index.ts is true, the request layer in src/lib/openrouter.ts excludes OpenRouter and Venice from all operations. This prevents any data from being sent to external APIs and disables telemetry to the G0DM0D3 service.

How does G0DM0D3 persist my LM Studio configuration?

The global Zustand store defined in src/store/index.ts serializes the localUrl and localOnly states into localStorage whenever they change. This ensures your LM Studio endpoint and mode preference persist across browser sessions and page reloads without requiring reconfiguration.

Can I use G0DM0D3 with other local inference servers besides LM Studio?

Yes. Any inference server implementing the OpenAI-compatible /v1/models and /v1/chat/completions specifications will work. Simply enter the server's base URL (e.g., http://localhost:5000/v1) into the Local Model URL field. The application treats all compatible endpoints identically, whether they come from LM Studio, Ollama, or custom implementations.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}