Basic Usage

Model Configuration

Each provider has its own configuration class.

OpenAI

Requires

Install "llmfy[openai]" and OPENAI_API_KEY environment variable.

from llmfy import OpenAIConfig

config = OpenAIConfig(temperature=0.7)

AWS Bedrock

Requires

Install "llmfy[boto3]" and AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_BEDROCK_REGION environment variables.

from llmfy import BedrockConfig

config = BedrockConfig(temperature=0.7)

Google AI

Requires

Install "llmfy[google-genai]" and GOOGLE_API_KEY environment variable.

from llmfy import GoogleAIConfig

config = GoogleAIConfig(temperature=0.7)

Initialize Model

OpenAI

from llmfy import OpenAIModel

llm = OpenAIModel(model="gpt-4o-mini", config=config)

AWS Bedrock

from llmfy import BedrockModel

llm = BedrockModel(model="amazon.nova-lite-v1:0", config=config)

Google AI

from llmfy import GoogleAIModel

llm = GoogleAIModel(model="gemini-2.5-flash-lite", config=config)

Create Agent

The LLMfy class is provider-agnostic — pass any initialized model:

from llmfy import LLMfy

agent = LLMfy(llm, system_message="You are a helpful assistant.")

Generate

Using Invoke

invoke accepts plain text or a list of Content objects:

response = agent.invoke("Hello")

print(f"\n>> {response.result.content}\n")

Using Chat

chat accepts a list of Message objects with roles:

from llmfy import Message, Role

messages = [Message(role=Role.USER, content="Hello")]

response = agent.chat(messages)

print(f"\n>> {response.result.content}\n")