gemini_client.py

import os
from dotenv import load_dotenv
from agents import OpenAIChatCompletionsModel, AsyncOpenAI

load_dotenv(override=True)
google_api_key = os.getenv("GOOGLE_API_KEY")
base_url="https://generativelanguage.googleapis.com/v1beta/openai/"

asyncllm = AsyncOpenAI(
    base_url=base_url,
    api_key=google_api_key
)

client = OpenAIChatCompletionsModel(
    openai_client=asyncllm,
     model="gemini-2.5-flash"
)