llmhub-cli 0.1.10

LLMHub is a lightweight management platform designed to streamline working with LLMs.

LLMHub CLI

LLMHub CLI is a command-line interface tool designed to manage and interact with various LLM (Large Language Model) servers. It allows you to start, stop, update, and manage LLM processes easily and efficiently.

Features

• Manage LLM servers
• Start, stop, and update LLM processes
• List available models and their statuses
• OpenAI-compatible API endpoints for completions and models
• Easily configurable via YAML files
• Supports different engines and quantization formats

Installation

You can install LLMHub CLI directly from PyPI:

pip install llmhub-cli

Usage

After installation, you can use the llmhub command to interact with the tool. Below are some example commands:

Start a Process

llmhub start MythoMax-L2-13B

Stop a Process

llmhub stop MythoMax-L2-13B

Update All Processes

llmhub update

List All Models

llmhub list-models

Check Status

llmhub status

Configuration

The configuration is handled via YAML files. You can place your config.yaml file in the ~/.llmhub/ directory or specify a custom path when initializing the ConfigManager.

Example Configuration

on_start:
  MythoMax-L2-13B:
    quant: Q5_K_M
    engine: llamacppserver
    context_size: [512, 1024, 2048]

port: 8080
enable_proxy: true
engine_port_min: 8081
engine_port_max: 10000

engines:
  llamacppserver:
    path: /path/to/llamacppserver
    arguments: --color -t 20 --parallel 2 --mlock --metrics --verbose
    model_flag: "-m"
    context_size_flag: "-c"
    port_flag: "--port"
    api_key_flag: "--api-key"
    file_types: [gguf]

API Endpoints

LLMHub CLI also provides OpenAI-compatible API endpoints:

• /v1/completions: Handle completion requests.
• /v1/chat/completions: Handle chat completion requests.
• /v1/models: List available models.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License.

Contact

For any questions or issues, please open an issue on the GitHub repository.