import logging
from openai import OpenAI
from typing import List, Generator, Optional

logging.basicConfig(level=logging.INFO)

def request_generation(
    api_key: str,
    api_base: str,
    message: str,
    system_prompt: str,
    model_name: str,
    chat_history: Optional[List[dict]] = None,
    temperature: float = 0.3,
    frequency_penalty: float = 0.0,
    presence_penalty: float = 0.0,
    max_new_tokens: int = 1024,
    tools: Optional[List[dict]] = None,
    tool_choice: Optional[str] = None,
) -> Generator[str, None, None]:
    """
    Sends a streaming chat request to an OpenAI-compatible backend using the official OpenAI client.
    Buffers output to improve LaTeX rendering.
    """
    client = OpenAI(api_key=api_key, base_url=api_base)

    messages = [{"role": "system", "content": system_prompt}]
    if chat_history:
        messages.extend(chat_history)
    messages.append({"role": "user", "content": message})

    request_args = {
        "model": model_name,
        "messages": messages,
        "temperature": temperature,
        "frequency_penalty": frequency_penalty,
        "presence_penalty": presence_penalty,
        "max_tokens": max_new_tokens,
        "stream": True,
    }

    if tools:
        request_args["tools"] = tools
    if tool_choice:
        request_args["tool_choice"] = tool_choice

    logging.info(f"[Gateway] Request to {api_base} | Model: {model_name}")

    try:
        stream = client.chat.completions.create(**request_args)

        collected = ""
        buffer = ""

        for chunk in stream:
            delta = chunk.choices[0].delta.content or ""
            collected += delta
            buffer += delta

            if "\n" in buffer or len(buffer) > 150:
                yield buffer
                buffer = ""

        if buffer:
            yield buffer

    except Exception as e:
        logging.exception("[Gateway] Streaming failed")
        yield f"Error: {e}"