LLM Tools

GPT Tester

GPT Tester is a lightweight macOS client (macOS 12+) designed to quickly test large language models via OpenAI-compatible APIs. It supports both OpenAI’s API and local LM Studio endpoints through built-in presets, making it useful for model comparison and prompt experimentation.

API Connection

Configurable endpoints:
One-click presets for OpenAI and LM Studio that automatically set the correct URLs.

Prompt Management

System message: define role/behavior of the model.
User message: enter queries or instructions.

Model and Parameters

Model selector: lists available models dynamically populated via /v1/models.
Temperature: shown as “not tunable” for some reasoning models.
Reasoning Effort: dropdown (e.g., medium) to adjust depth of reasoning.
Verbosity: dropdown (e.g., medium) to adjust level of detail in responses.

Download GPT Tester

LM Server

LM Server is a lightweight web service layer designed to expose large language models (LLMs) running on LM Studio to multiple users across a local network (LAN). By combining LM Studio’s inference engine with LM Server’s browser-based interface, it enables one high-performance machine to provide a simple, multi-user LLM service to any device with a browser.

Integration with LM Studio

Back-end powered by LM Studio: LM Server connects to LM Studio’s OpenAI-compatible API endpoints.
Model selection: Models available in LM Studio are surfaced in LM Server, and an administrator can choose which model to serve.

LAN-Based Access

Clients on the same network can connect via their browser using the server’s LAN IP.
No need for GPU or model installation on client devices.

Per-User Sessions

Each connected browser session has its own conversation history.
Multiple users can chat with the same model simultaneously without interference.

Streaming Mode

Supports streaming responses, so replies are generated token by token in real time, improving interactivity.

Customizable Experience

System prompt customization: Define the model’s role, style, or behavior globally.
User-facing message customization: Adjust welcome messages and interaction hints for end users.
Allows easy creation of custom chatbot services tailored to a team, classroom, or small organization.

Download LM Server

FM Server

FM Server is a lightweight service layer, similar in design to LM Server, but instead of connecting to LM Studio, it leverages the built-in Foundation Model in macOS Tahoe 26. This makes it possible to run Apple’s native LLM as the backend while still providing a browser-based, multi-user chat interface across a local network.

Integration with macOS Foundation Model

• Back-end powered by Apple’s Foundation Model (macOS Tahoe 26) instead of LM Studio.

• Provides access to Apple’s optimized, on-device LLM without requiring third-party model downloads.

• Administrators can select the model to serve directly from FM Server.

LAN-Based Web Access

• Clients connect from any device on the same LAN via a standard web browser.

• No installation or heavy compute required on client devices.

Independent User Sessions

• Each browser session maintains its own conversation history.

• Multiple users can interact with the Foundation Model in parallel, without interference.

Streaming Responses

• Supports real-time streaming output, allowing users to see text generated token by token.

• Improves responsiveness and user experience, especially with longer outputs.

Customization

• System prompts: Define the model’s role, style, or constraints globally.

• User-facing messages: Customize welcome text and interaction instructions for clients.

• Enables deployment of custom conversational services using Apple’s native models.

Download FM Server

These are tools intended for research and experimental purposes.

All applications are provided as-is.

The developers bear no responsibility for the outcomes of their operation; please use them at your own risk.

Page updated

Google Sites

Report abuse