custom_openai

custom_openai is a drop-in replacement for the official OpenAI Python SDK, enabling you to easily customize, standardize, and enrich the API responses for your own post-processing, logging, RAG pipelines, or monitoring needs. It supports both synchronous and asynchronous usage, with robust support for streaming and complete compatibility with the OpenAI API.

🚀 Features

• Plug-and-play: Fully compatible with the OpenAI client interface. Just swap your import and go.
• Custom response fields: Every .create() call (sync/async/streaming) includes a .flashquery attribute for easy extraction of processed or enriched response data.
• Streaming support: Automatically injects your custom data into the final item in async generator streams.
• No monkey-patching: Cleanly extends OpenAI’s classes without risky global side effects.

📦 Installation

pip install flashquery

🔥 Quickstart

Synchronous Example

from flashquery.client import CustomLangchainClient

client = CustomLangchainClient(
    provider="openai",
    model="gpt-4o-mini",
    temperature=0,
    api_key="sk-..."
)

response = client.generate([{"role": "user", "content": "Say hello?"}])

print(response.flashquery)  # Your custom field!

Streaming Example

import asyncio
from flashquery.client import CustomLangchainClient

async def main():
    client = CustomLangchainClient(
        provider="openai",
        model="gpt-4o-mini",
        temperature=0,
        api_key="sk-..."
    )
    response = client.astream([{"role": "user", "content": "Oi, tudo bem?"}])

    async for chunk in response:
        print("Chunk:", chunk.flashquery)

asyncio.run(main())

⚙️ How it Works

• CustomOpenAIClient and CustomAsyncOpenAIClient inherit from the official OpenAI clients, overriding .create() methods of responses and chat.completions.
• After each response is created, a new .flashquery attribute is attached.
• In streaming mode, .flashquery is set on the last chunk yielded from the generator.

🤝 Contributing

Pull requests are welcome! For major changes, please open an issue first to discuss what you would like to change.

📄 License

MIT License

Need help? Open an issue or contribute on GitHub!