Google (Gemini)

The GoogleLLMService integrates Google AI's Gemini models for powerful conversational reasoning. It supports the standard LLM features along with Gemini-specific capabilities like "Thinking" mode for deep reasoning.

Installation

To use Google Gemini, install the required dependencies:

pip install "piopiy-ai[google]"

Prerequisites

• A Google AI Studio API key (Get yours here).
• Set your API key in your environment:

  export GOOGLE_API_KEY="your_api_key_here"

Configuration

GoogleLLMService Parameters

Parameter	Type	Default	Description
api_key	str	Required	Your Google AI API key.
model	str	"gemini-2.5-flash"	Gemini model ID.
params	InputParams	InputParams()	Model generation settings.
system_instruction	str	None	System prompt for the model.

InputParams

Parameter	Type	Default	Description
max_tokens	int	4096	Maximum generation length.
temperature	float	None	Sampling temperature.
thinking	ThinkingConfig	None	Configuration for "Thinking" models.

Usage

Basic Setup

import os
from piopiy.services.google.llm import GoogleLLMService

llm = GoogleLLMService(
    api_key=os.getenv("GOOGLE_API_KEY"),
    model="gemini-2.5-flash"
)

Using Thinking Mode

from piopiy.services.google.llm import GoogleLLMService

llm = GoogleLLMService(
    api_key=os.getenv("GOOGLE_API_KEY"),
    model="gemini-2.0-pro-exp",
    params=GoogleLLMService.InputParams(
        thinking=GoogleLLMService.ThinkingConfig(
            thinking_budget=1024
        )
    )
)

Notes

• Thinking Budget: For high-reasoning models (like Gemini 2.5 Pro), you can set a thinking_budget (in tokens) to control the depth of the reasoning process.
• Multimodal: Gemini models natively support images and audio in the context.