Environment setup

API Keys

Set the API key for each provider you plan to use:

Provider	Environment Variable
OpenAI	OPENAI_API_KEY
Anthropic	ANTHROPIC_API_KEY
Google Gemini	GEMINI_API_KEY
DeepSeek	DEEPSEEK_API_KEY
Cohere	COHERE_API_KEY
Perplexity	PERPLEXITY_API_KEY
OpenRouter	OPENROUTER_API_KEY
Moonshot	MOONSHOT_API_KEY
Voyage AI (embeddings)	VOYAGE_API_KEY

Note: Ollama and Hugging Face local models don't require API keys.

Using a .env File

Create a .env file in your project root:

OPENAI_API_KEY=sk-your-key-here
ANTHROPIC_API_KEY=sk-ant-your-key-here
GEMINI_API_KEY=your-key-here

SimplerLLM loads the .env file automatically. No extra code needed.

Passing Keys Directly

You can also pass API keys directly when creating an LLM instance:

llm = LLM.create(
    provider=LLMProvider.OPENAI,
    model_name="gpt-4o",
    api_key="sk-your-key-here"
)

This overrides any environment variable.

Global Settings

Optional settings that apply to all providers:

MAX_RETRIES=3      # Number of retry attempts on failure (default: 3)
RETRY_DELAY=2      # Seconds between retries (default: 2)

Local Models

Ollama

OLLAMA_URL=http://localhost:11434/   # Ollama server URL (default)
OLLAMA_TIMEOUT=120                    # Request timeout in seconds

Hugging Face Local

HF_DEVICE=auto           # "cuda", "cpu", "mps", or "auto"
HF_TORCH_DTYPE=auto      # "float16", "bfloat16", "float32", or "auto"
HF_TIMEOUT=300           # Generation timeout in seconds
HF_LOAD_IN_4BIT=false    # Enable 4-bit quantization
HF_LOAD_IN_8BIT=false    # Enable 8-bit quantization

OpenRouter

Optional metadata for request tracking:

OPENROUTER_API_KEY=your-key-here
OPENROUTER_SITE_URL=https://yoursite.com
OPENROUTER_SITE_NAME=Your App Name