kiarina-lib-openai

English | 日本語

Dependencies

The OpenAI Python SDK is not included as a dependency. Add it to the application that creates the client.

Installation

pip install kiarina-lib-openai

Features

• Configuring an OpenAI Client Generate OpenAI SDK client arguments from settings.
• Managing Multiple Configurations Switch between named settings for multiple projects or environments.
• Loading Environment Variables Load environment variables with the KIARINA_LIB_OPENAI_ prefix.
• Using Compatible APIs Set an OpenAI-compatible API endpoint with base_url.
• Protecting API Keys Store the API key as SecretStr.

Configuring an OpenAI Client

to_client_kwargs() converts only configured values to OpenAI SDK argument names.

from openai import OpenAI

from kiarina.lib.openai import OpenAISettings

settings = OpenAISettings(api_key="sk-...")
client = OpenAI(**settings.to_client_kwargs())

organization_id becomes organization. When api_key, organization_id, and base_url are all None, the method returns an empty dict.

Managing Multiple Configurations

settings_manager uses multiple-settings mode. Place named settings under configs.

kiarina.lib.openai:
  default: production
  configs:
    development:
      api_key: sk-development
    production:
      api_key: sk-production
      organization_id: org-production

import yaml
from pydantic_settings_manager import load_user_configs

from kiarina.lib.openai import settings_manager

with open("config.yaml", encoding="utf-8") as file:
    load_user_configs(yaml.safe_load(file) or {})

settings = settings_manager.get_settings("production")

To configure only this package directly, assign settings_manager.user_config.

from kiarina.lib.openai import settings_manager

settings_manager.user_config = {
    "default": "development",
    "configs": {
        "development": {"api_key": "sk-development"},
        "production": {
            "api_key": "sk-production",
            "organization_id": "org-production",
        },
    },
}

Loading Environment Variables

Load a single configuration from environment variables.

export KIARINA_LIB_OPENAI_API_KEY="sk-..."
export KIARINA_LIB_OPENAI_ORGANIZATION_ID="org-..."
export KIARINA_LIB_OPENAI_BASE_URL="https://api.openai.com/v1"

from kiarina.lib.openai import OpenAISettings

settings = OpenAISettings()

Using Compatible APIs

from openai import OpenAI

from kiarina.lib.openai import OpenAISettings

settings = OpenAISettings(
    api_key="compatible-api-key",
    base_url="https://example.com/v1",
)
client = OpenAI(**settings.to_client_kwargs())

Protecting API Keys

to_client_kwargs() returns the secret value of api_key so it can be passed to the OpenAI SDK. Do not log the returned dictionary.

API Reference

kiarina.lib.openai

from kiarina.lib.openai import (
    OpenAISettings,
    settings_manager,
)

OpenAISettings

class OpenAISettings(BaseSettings):
    def __init__(
        self,
        *,
        api_key: SecretStr | None = None,
        organization_id: str | None = None,
        base_url: str | None = None,
    ) -> None: ...

    def to_client_kwargs(self) -> dict[str, Any]: ...

OpenAI client settings.

Fields

• api_key (SecretStr | None): OpenAI API key.
• organization_id (str | None): OpenAI organization ID.
• base_url (str | None): Base URL for the OpenAI API.

to_client_kwargs

Converts fields whose values are not None into OpenAI SDK client arguments.

Returns

• dict[str, Any]: A dictionary containing configured values among api_key, organization, and base_url.

settings_manager

settings_manager: SettingsManager[OpenAISettings]

Global instance that manages named OpenAI client settings.