LangCommand

LangCommand is a local inference command-line tool powered by llama.cpp that transforms natural language descriptions into executable shell commands.

• Natural Language Command Generation: Simply describe the command you need in plain language, and LangCommand will generate the appropriate shell command for you.
• Customizable Models: Choose from a variety of preconfigured models or provide your own for custom workflows.
• Multiple Modes:
  • loop mode: Continues to choose and execute commands.
  • exit mode: Executes a single command and exits.
• Command Explanation: Optionally provides an explanation of the generated command.
• Tracing: Enable model tracing to debug and track the execution process.

Install

To build LangCommand from source, use the following command:

make lac

Usage

To use LangCommand, provide a natural language description:

lac "your prompt"

You can run lac without arguments to explore options:

--------------------------------- LangCommand params ----------------------------------
-h,    --help, --usage                  Print LangCommand usage
--setup                                 Set up your LangCommand model: choose or customize
--show-args                             Show arguments you saved
--no-explanation                        Disable command explanation
--mode {loop,exit}                      Select the mode of operation.
                                                - loop: Continues to choose and execute commands indefinitely.
                                                - exit: Executes a single command and then stops the program.
--model-help, --model-usage             Print LangCommand default model arguments
--trace                                 Enable tracing for the execution of the default model

Supported Models

LangCommand comes with support for the following preconfigured models:

• Qwen2.5-Coder-7B
• Qwen2.5-7B
• Llama-3.2-3B
• Llama-3.2-1B
• Codellama-13B

You can also provide your own model with custom system prompts.

Troubleshooting

ggml_metal_graph_compute: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)

This typically occurs when the system does not have enough memory to load and run the selected model. To resolve, consider using a smaller model.

Community & Contact

Connect to discuss LangCommand, share feedback, and get support from other users:

Join the LangCommand Discord group

For any questions or inquiries, feel free to reach out via email:

Email me