universal-llm 0.1.0

Minimal universal LLM client - swap providers with one env var

Universal LLM Client

Swap LLM providers with one env var. No code changes.

from universal_llm import Settings, get_client

settings = Settings(provider="anthropic", model="claude-3-haiku")
client = get_client(settings)

resp = client.chat_sync([
    {"role": "user", "content": "Explain the CAP theorem in 2 sentences."}
])
print(resp)

Installation

pip install universal-llm[all]

Or install only what you need:

• pip install universal-llm[openai]
• pip install universal-llm[anthropic]
• pip install universal-llm[google]
• pip install universal-llm[ollama]

Configuration

Set via environment variables:

• LLM_PROVIDER - openai, anthropic, google, or ollama
• LLM_MODEL - model name
• LLM_API_KEY - API key (not required for Ollama)
• LLM_BASE_URL - base URL (Ollama defaults to http://localhost:11434/v1)
• LLM_TEMPERATURE - temperature (default: 0.2)
• LLM_TIMEOUT - timeout in seconds (default: 60)