llms2

A Python package for managing multiple LLM profiles and API interactions.

Installation

Install the package using pip:

pip install -e .

• or pip install git+https://github.com/SanBingYouYong/llms2.git

Or with uv:

uv pip install -e .

• or uv add git+https://github.com/SanBingYouYong/llms2.git

Configuration

Create a llms.yml file with your LLM profiles in your project root (see an example in this project root):

profiles:
  example_profile:
    url: https://api.example.com
    api_key_env: EXAMPLE_API_KEY
    model: model-name
    vision: false
    accepts_sys_msg: true
    request_params:
      temperature: 1

Set the corresponding environment variables in a .env file or export them:

EXAMPLE_API_KEY=your_api_key_here

Usage

from llms2 import LLMS

# Initialize with configuration file (default: llms.yml, you may as well call it differently and pass in the path)
llms = LLMS()

# Prepare messages
messages = llms.prepare_messages(
    prompt="Hello, how are you?",
    system_msg="When the user asks you how are you, always reply with hello world."
)

# Query an LLM profile
response = llms.query("deepseek-v3.2", messages)
print(response)

# Save conversation history
messages.append({"role": "assistant", "content": response})
llms.save_messages("deepseek-v3.2", messages, "conversation.yml")

# Load conversation history
loaded_messages = llms.load_messages("conversation.yml")

Features

• Support for multiple LLM profiles with different configurations
• Vision model support for image inputs
• Conversation history management (save/load)
• System message support

Development Notes

• Streaming and visualization features: no longer supported comparing to llms
• "Distributed" chat history dump comparing to llms