lolm 0.1.0

Lightweight Orchestrated LLM Manager - Multi-provider LLM configuration and management

LOLM - Lightweight Orchestrated LLM Manager

Reusable LLM configuration and management package for Python projects.

Multi-provider support with automatic fallback, priority routing, and unified configuration.

✨ Features

• 🔄 Multi-provider support - OpenRouter, Ollama, Groq, Together, LiteLLM
• ⚡ Automatic fallback - Seamless provider switching on failure
• 🎯 Priority routing - Configure provider order by priority
• 🔧 Unified config - .env, litellm_config.yaml, ~/.lolm/
• 🖥️ CLI interface - Similar to reclapp llm

🚀 Installation

pip install lolm

Or with optional dependencies:

pip install lolm[full]      # All providers
pip install lolm[ollama]    # Ollama support
pip install lolm[litellm]   # LiteLLM support

📖 Quick Start

CLI

# Show provider status
lolm status

# Set default provider
lolm set-provider openrouter

# Set model
lolm set-model openrouter nvidia/nemotron-3-nano-30b-a3b:free

# Manage API keys
lolm key set openrouter YOUR_API_KEY

# Test generation
lolm test

Python API

from lolm import get_client, LLMManager

# Simple usage
client = get_client()
response = client.generate("Explain this code")

# With specific provider
client = get_client(provider='openrouter')
response = client.generate("Hello!", system="You are helpful")

# With manager for fallback
manager = LLMManager()
manager.initialize()

response = manager.generate_with_fallback(
    "Generate code",
    providers=['openrouter', 'groq', 'ollama']
)

⚙️ Configuration

Environment Variables (.env)

# API Keys
OPENROUTER_API_KEY=sk-or-v1-...
GROQ_API_KEY=gsk_...
TOGETHER_API_KEY=...

# Default provider
LLM_PROVIDER=auto

# Model overrides
OPENROUTER_MODEL=nvidia/nemotron-3-nano-30b-a3b:free
OLLAMA_MODEL=qwen2.5-coder:14b

litellm_config.yaml

model_list:
  - model_name: code-analyzer
    litellm_params:
      model: ollama/qwen2.5-coder:7b
      api_base: http://localhost:11434
    priority: 10

router_settings:
  routing_strategy: simple-shuffle
  num_retries: 3

🖥️ CLI Reference

Command	Description
lolm status	Show provider status
lolm set-provider PROVIDER	Set default provider
lolm set-model PROVIDER MODEL	Set model for provider
lolm key set PROVIDER KEY	Set API key
lolm key show	Show configured keys
lolm models [PROVIDER]	List recommended models
lolm test [--provider P]	Test LLM generation
lolm config show	Show configuration
lolm priority set-provider P N	Set provider priority
lolm priority set-mode MODE	Set priority mode

🔌 Supported Providers

Provider	Type	Free Tier	Default Model
OpenRouter	Cloud	✓	nvidia/nemotron-3-nano-30b-a3b:free
Ollama	Local	✓	qwen2.5-coder:14b
Groq	Cloud	✓	llama-3.1-70b-versatile
Together	Cloud	-	Qwen/Qwen2.5-Coder-32B-Instruct
LiteLLM	Universal	-	gpt-4

🧪 Development

# Install dev dependencies
make install-dev

# Run tests
make test

# Format code
make format

# Lint
make lint

# Build package
make build

# Publish to PyPI
make publish

📄 License

Apache 2.0 License - see LICENSE for details.

🔗 Links

• Documentation
• PyPI
• GitHub
• Issues