app.py

import gradio as gr
from huggingface_hub import InferenceClient

"""
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""
# client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")


from google.cloud import storage
from google.oauth2 import service_account
import json
import os
import requests

# upload image to google cloud storage
def upload_file_to_gcs_blob(file):

    google_creds = os.environ.get("GOOGLE_APPLICATION_CREDENTIALS_JSON")

    creds_json = json.loads(google_creds)
    credentials = service_account.Credentials.from_service_account_info(creds_json)

    # Google Cloud credentials
    storage_client = storage.Client(credentials=credentials, project=creds_json['project_id'])

    bucket_name=os.environ.get('bucket_name')
    bucket = storage_client.bucket(bucket_name)
    
    destination_blob_name = os.path.basename(file)
    blob = bucket.blob(destination_blob_name)

    blob.upload_from_filename(file)

    public_url = blob.public_url
    
    return public_url


from PIL import Image

def is_image(file_path):
    try:
        Image.open(file_path)
        return True
    except IOError:
        return False

def supbase_insert(user_message,response_content,messages,response):
    from supabase import create_client, Client
    url = os.environ.get('supabase_url')
    key = os.environ.get('supabase_key')
    supabase = create_client(url, key)
    data, count = supabase.table('messages').insert({"user_message": user_message, "response_content": response_content,"messages":messages,"response":response}).execute()


# def respond(
#     message,
#     history: list[tuple[str, str]],
#     system_message,
#     max_tokens,
#     temperature,
#     top_p,
# ):
#     messages = [{"role": "system", "content": system_message}]

#     for val in history:
#         if val[0]:
#             messages.append({"role": "user", "content": val[0]})
#         if val[1]:
#             messages.append({"role": "assistant", "content": val[1]})

#     messages.append({"role": "user", "content": message})

#     response = ""

#     for message in client.chat_completion(
#         messages,
#         max_tokens=max_tokens,
#         stream=True,
#         temperature=temperature,
#         top_p=top_p,
#     ):
#         token = message.choices[0].delta.content

#         response += token
#         yield response

def get_completion(message,history):
    if message["text"].strip() == "" and not message["files"]:
        raise gr.Error("Please input a query and optionally image(s).")
    
    if message["text"].strip() == "" and message["files"]:
        raise gr.Error("Please input a text query along the image(s).")
    
    text = message['text']
    user_message = [
        {"type": "text", "text": text},
    ]
    if message['files']:
        file = message['files'][0]
        public_url = upload_file_to_gcs_blob(file)
        if is_image(file): # only support image file now
            content_image = {
                "type": "image_url",
                "image_url": {
                    "url": public_url,
                },}
            user_message.append(content_image)
        else:
            raise gr.Error("Only support image files now.")

    history_openai_format = []
    for human, assistant in history:
        # check if there is image info in the history message
        if isinstance(human, tuple):
            continue
        history_openai_format.append({"role": "user", "content": human })
        history_openai_format.append({"role": "assistant", "content":assistant})
    history_openai_format.append({"role": "user", "content": user_message})
    # print(history_openai_format)
    
    system_message = '''You are GPT-4o("o" for omni), OpenAI's new flagship model that can reason across audio, vision, and text in real time. 
    GPT-4o matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster. 
    GPT-4o is especially better at vision and audio understanding compared to existing models.
    GPT-4o's text and image capabilities are avaliable for users now. More capabilities like audio, and video will be rolled out iteratively in the future.
    '''

    
    # headers
    openai_api_key = os.environ.get('openai_api_key')
    headers = {
      'Authorization': f'Bearer {openai_api_key}'
    }

    temperature = 0.7
    max_tokens = 2048

    init_message = [{"role": "system", "content": system_message}]
    messages = init_message + history_openai_format[-5:] #system message + latest 2 round dialogues + user input
    print(messages)
    # request body
    data = {
        'model': 'gpt-4o',  # we use gpt-4o here
        'messages': messages,
        'temperature':temperature, 
        'max_tokens':max_tokens,
        # 'stream':True，
    }

    # get response
    response = requests.post('https://burn.hair/v1/chat/completions', headers=headers, json=data)
    response_data = response.json()
    print(response_data)
    print('-----------------------------------\n')
    if 'error' in response_data:
        response_content = response_data['error']['message']
    else:
        response_content = response_data['choices'][0]['message']['content']
        usage = response_data['usage']

    supbase_insert(user_message,response_content,messages,response_data)
    
    return response_content



"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    get_completion,
    multimodal=True,
    # additional_inputs=[
    #     gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
    #     gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
    #     gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
    # ],
)

demo.queue(max_size = 10)


if __name__ == "__main__":
    demo.launch()