utils/caller/llm_client.py

import base64
from openai import Client
from openai.types.chat.chat_completion import ChatCompletion
from gradio_client.utils import is_http_url_like
import magic
from pydantic import BaseModel
from ..learner.learner import DefaultTool
from typing import Any
from copy import deepcopy
import json

def get_client(api_key: str | None = None, **kwargs):
    return Client(
        api_key=api_key,
        **kwargs,
    )

def encode_image(image_path:str):
    with open(image_path, "rb") as image_file:
        base64_image = base64.b64encode(image_file.read()).decode("utf-8")
    mime = magic.Magic(mime=True)
    mime_type = mime.from_file(image_path)
    return f"data:{mime_type};base64,{base64_image}"

def image_to_one_of_content(
    image_path:str,
    detail:str="auto",
):
    url = (
        image_path
        if is_http_url_like(image_path) else
        encode_image(image_path)
    )

    return {
        "type":"image_url",
        "image_url":{
            "url":url,
            "detail":detail,
        }
    }

def audio_to_content(
    data:str,
    format:str,
):
    return {
        "type":"input_audio",
        "input_audio":{
            "data": data,
            "format": format,
        }
    }

def chat_completions(
    messages: list,
    model:str,
    *,
    client : Client | None = None,
    **kwargs,
)->ChatCompletion:
    client = client or get_client()

    response = client.chat.completions.create(
        model=model,
        messages=messages,
        **kwargs,
    )

    return response

def tts(
    input,
    *,
    client : Client | None = None,
    voice="fable",
    model="tts-1"
):
    client = client or get_client()

    response = client.audio.speech.create(
        model=model,
        voice=voice,
        input=input,
    )

    return response