LLM Interface

Creating an LLM Instance

from SimplerLLM.language.llm import LLM, LLMProvider

llm = LLM.create(provider=LLMProvider.OPENAI, model_name="gpt-4o")
response = llm.generate_response(prompt="What is Python?")

Parameter	Type	Default	Description
provider	LLMProvider	OPENAI	The LLM provider to use
model_name	str	"gpt-4o-mini"	Model identifier
temperature	float	0.7	Sampling temperature (0-2)
top_p	float	1.0	Nucleus sampling (0-1)
api_key	str	None	API key (defaults to environment variable)

Providers

Provider	Enum Value	Environment Variable
OpenAI	LLMProvider.OPENAI	OPENAI_API_KEY
Google Gemini	LLMProvider.GEMINI	GEMINI_API_KEY
Anthropic	LLMProvider.ANTHROPIC	ANTHROPIC_API_KEY
DeepSeek	LLMProvider.DEEPSEEK	DEEPSEEK_API_KEY
Ollama	LLMProvider.OLLAMA	— (local)
OpenRouter	LLMProvider.OPENROUTER	OPENROUTER_API_KEY
Cohere	LLMProvider.COHERE	COHERE_API_KEY
Perplexity	LLMProvider.PERPLEXITY	PERPLEXITY_API_KEY
Moonshot	LLMProvider.MOONSHOT	MOONSHOT_API_KEY
HuggingFace Local	LLMProvider.HUGGING_FACE_LOCAL	— (local)

# Example: Using Anthropic
llm = LLM.create(provider=LLMProvider.ANTHROPIC, model_name="claude-sonnet-4-20250514")

# Example: Using local Ollama
llm = LLM.create(provider=LLMProvider.OLLAMA, model_name="llama3")

Generation Methods

generate_response

Generate text synchronously.

# Using a prompt
response = llm.generate_response(prompt="Explain quantum computing")

# Using chat messages
response = llm.generate_response(
    messages=[
        {"role": "user", "content": "What is AI?"},
        {"role": "assistant", "content": "AI is artificial intelligence..."},
        {"role": "user", "content": "Give me an example"}
    ]
)

generate_response_async

Generate text asynchronously.

import asyncio

async def main():
    response = await llm.generate_response_async(prompt="Explain machine learning")
    print(response)

asyncio.run(main())

Core Parameters

Parameter	Type	Default	Description
prompt	str	None	Text input (use this OR messages)
messages	list[dict]	None	Chat history (use this OR prompt)
system_prompt	str	"You are a helpful AI Assistant"	System instruction
temperature	float	0.7	Sampling temperature
max_tokens	int	300	Maximum output tokens
top_p	float	1.0	Nucleus sampling
json_mode	bool	False	Force JSON output
full_response	bool	False	Return detailed response object
images	list[str]	None	List of image URLs or local file paths for vision
detail	str	"auto"	Image detail level: "low", "high", or "auto"
web_search	bool	False	Enable web search

Vision / Image Input

Pass images to analyze visual content:

# Using an image URL
response = llm.generate_response(
    prompt="Describe this image",
    images=["https://example.com/photo.jpg"]
)

# Using a local file
response = llm.generate_response(
    prompt="What objects are in these images?",
    images=["path/to/image1.png", "path/to/image2.jpg"],
    detail="high"  # "low", "high", or "auto"
)

Provider	Vision Support
OpenAI	Supported
Anthropic	Supported
Gemini	Supported
Ollama	Supported (llava, llama3.2-vision)
Cohere	Supported
DeepSeek	Supported

JSON Mode

response = llm.generate_response(
    prompt="List 3 programming languages as JSON",
    json_mode=True
)

Full Response

Get detailed metadata including token counts and timing.

response = llm.generate_response(
    prompt="What is Python?",
    full_response=True
)

print(response.generated_text)
print(f"Input tokens: {response.input_token_count}")
print(f"Output tokens: {response.output_token_count}")
print(f"Time: {response.process_time}s")

Note: Use prompt for single queries or messages for conversations. Do not use both.