lmsp 0.5.0

A command-line interface for sending prompts to LM Studio loaded models

lmsp - LM Studio Prompt CLI

A simple command-line interface for sending prompts to LM Studio loaded models.

Features

• Send prompts to locally loaded LM Studio models
• Uses the first loaded model by default (or specify with -m)
• Requires pre-loaded models: Models must be loaded using lms load <model> or LM Studio desktop app
• Support for piping input from other commands
• Verbose logging with -v flag for debugging
• Simple and fast command-line interface

Installation

Quick Install from PyPI (Recommended)

# Install globally with pip
pip install lmsp

# Or install globally with uv (recommended)
uv tool install lmsp

Alternative Installation Methods

Install from source

# Using uv tool (recommended - installs globally)
uv tool install git+https://github.com/kmlawson/lmsp.git

# Or clone and install locally
git clone https://github.com/kmlawson/lmsp.git
cd lmsp
uv tool install .

Install in virtual environment

# Using uv
uv venv
source .venv/bin/activate
uv pip install lmsp

# Or using pip
python -m venv venv
source venv/bin/activate
pip install lmsp

Development installation

# Clone and install in development mode
git clone https://github.com/kmlawson/lmsp.git
cd lmsp
uv pip install -e .  # or pip install -e .

Configuration

lmsp supports a configuration file to set default values for command-line options. The configuration file is located at ~/.lmsp-config and is automatically created with default values when you first run lmsp.

Configuration File Format

The configuration file uses JSON format:

{
  "model": null,
  "port": 1234,
  "pipe_mode": "replace",
  "wait": false,
  "stats": false,
  "plain": false,
  "verbose": false
}

Configuration Options

• model: Default model to use (null means use first loaded model)
• port: Default LM Studio server port (1234)
• pipe_mode: How to handle piped input ("replace", "append", or "prepend")
• wait: Disable streaming by default (false)
• stats: Show response statistics by default (false)
• plain: Disable markdown formatting by default (false)
• verbose: Enable verbose logging by default (false)

Example Custom Configuration

{
  "model": "google/gemma-3n-e4b",
  "port": 1234,
  "pipe_mode": "append",
  "wait": true,
  "stats": true,
  "plain": false,
  "verbose": false
}

This configuration would:

• Use "google/gemma-3n-e4b" as the default model
• Wait for complete responses (no streaming) and beautify markdown output
• Show response statistics by default
• Append piped content to prompts

Command-line arguments always override configuration file settings.

Usage

Prerequisites

Before using lmsp, you need to load a model:

# Load a model using lms command
lms load google/gemma-3n-e4b

# Or use LM Studio desktop app to load a model

Basic usage

lmsp "What is the capital of France?"

Specify a model

# Use a specific model (must be already loaded)
lmsp -m llama-3.2-1b-instruct "Explain quantum computing"

# Enable verbose logging for debugging
lmsp -v -m google/gemma-3n-e4b "What is AI?"

Pipe input

# Simple piping - replaces the prompt
cat document.txt | lmsp

# Combine prompt with piped content (default appends)
cat document.txt | lmsp "Summarize this document:"

# Control how piped input is combined
cat context.txt | lmsp "Answer based on context:" --pipe-mode prepend
cat document.txt | lmsp "Summarize:" --pipe-mode append

# Real example: Translate a text to English
cat tests/testdata/test-text.md | lmsp "Please translate the following text to English:"

Check loaded models

# List currently loaded models
lmsp --list-models

# List all available models (not loaded)
lms ls

Check server status

lmsp --check-server

Get help

lmsp --help
# or lmsp -h

Security Considerations

When using lmsp, please be aware of the following security considerations:

Piped Content

• Be cautious about what content you pipe to lmsp. The piped content is directly appended or prepended to your prompt without sanitization.
• Avoid piping untrusted content or files from unknown sources
• Be especially careful when piping content that might contain prompt injection attempts or malicious instructions
• Example of what to avoid:

  # Don't pipe untrusted user input or files
  cat untrusted_user_file.txt | lmsp "Summarize this:"
  

Model Selection

• Only use trusted models that you have intentionally loaded into LM Studio
• Be aware that models will execute the prompts you send, including any piped content

Local Usage

• lmsp is designed for local use with your own LM Studio instance
• It connects to localhost only and does not expose any network services

Prerequisites

1. LM Studio must be installed
2. The LM Studio server must be running (lms server start)
3. At least one model must be loaded (lms load <model>)

Running Tests

python -m unittest tests.test_lmsp -v

Planned Features

• Ability to attach images with -a flag for multi-modal models
• Ability to continue from last prompt
• Enhanced piping support for documents