OpenAI LLM

The OpenAILLMService provides conversational reasoning and response generation using OpenAI's GPT models. It supports streaming completions, tool usage (function calling), and vision capabilities.

Installation

To use OpenAI, install the required dependencies:

pip install "piopiy-ai[openai]"

Prerequisites

• An OpenAI account and API key (Get yours here).
• Set your API key in your environment:

  export OPENAI_API_KEY="your_api_key_here"

Configuration

OpenAILLMService Parameters

Parameter	Type	Default	Description
model	str	"gpt-4o"	The OpenAI model name to use.
api_key	str	None	OpenAI API key (defaults to env var).
base_url	str	None	Custom base URL for the API.
params	InputParams	InputParams()	Advanced sampling and token settings.

InputParams

Parameter	Type	Default	Description
temperature	float	1.0	Sampling temperature (0.0 to 2.0).
max_tokens	int	None	Maximum tokens in response.
top_p	float	1.0	Nucleus sampling parameter.
presence_penalty	float	0.0	Penalty for new topics.
frequency_penalty	float	0.0	Penalty for repeat tokens.

Usage

Basic Setup

import os
from piopiy.services.openai.llm import OpenAILLMService

llm = OpenAILLMService(
    api_key=os.getenv("OPENAI_API_KEY"),
    model="gpt-4o"
)

With Tool Usage

from piopiy.services.openai.llm import OpenAILLMService

llm = OpenAILLMService(
    api_key=os.getenv("OPENAI_API_KEY"),
    model="gpt-4o"
)

# Tools are typically registered on the VoiceAgent
# voice_agent.register_tool("get_weather", weather_handler)

Notes

• Model Choice: gpt-4o is recommended for the best balance of speed and intelligence.
• Streaming: Tokens are streamed to the TTS service as they are generated, minimizing response latency.