app.py

import time
import base64
from openai import OpenAI
import gradio
import google.generativeai as genai


def generation(prompt_text):
    a_key = "QUl6YVN5RGRJZXlsSnVUOERWMndlbXBfeE8wWUpmZEl4OFZ4S1FJ"
    genai.configure(api_key=str(base64.b64decode(a_key), encoding="utf-8"))

    generation_config = {"temperature": 0.9, "top_p": 1, "top_k": 1, "max_output_tokens": 2048}
    safety_settings = [
      {"category": "HARM_CATEGORY_HARASSMENT", "threshold": "BLOCK_MEDIUM_AND_ABOVE"},
      {"category": "HARM_CATEGORY_HATE_SPEECH", "threshold": "BLOCK_MEDIUM_AND_ABOVE"},
      {"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT", "threshold": "BLOCK_MEDIUM_AND_ABOVE"},
      {"category": "HARM_CATEGORY_DANGEROUS_CONTENT", "threshold": "BLOCK_MEDIUM_AND_ABOVE"}]

    model = genai.GenerativeModel(model_name="gemini-pro", generation_config=generation_config, safety_settings=safety_settings)

    response = model.generate_content(prompt_text)
    return response


chatgpt_model_type_normal = "gpt-3.5-turbo-instruct：普通对话型"
chatgpt_model_type_memory = "gpt-3.5-turbo-16k：记忆对话型"
chatgpt_model_type_gemini = "这是Google旗下原生多模态大模型 Gemini google-gemini-pro：普通对话型"
conversation_history = []
prompt_parts = [""]
def chatgpt(input_text, model_type, field):
    a_key = "c2stRFo3ZzRadXlzOEhMd2RWZnFYVWJUM0JsYmtGSkFPTm9PNHdPMmJZTTc5Tm5DTVQ5"
    client = OpenAI(api_key=str(base64.b64decode(a_key), encoding="utf-8"))

    if field != "":
        input_text = input_text + "\n" + "假设你扮演的人物角色或你专注的学术领域为：" + field

    if model_type == chatgpt_model_type_normal:
        response = client.completions.create(
            model="gpt-3.5-turbo-instruct",
            temperature=0.25,
            max_tokens=2500,
            frequency_penalty=0.5,
            presence_penalty=0.5,
            top_p=0.88,
            prompt=input_text
        )
        response_text = response.choices[0].text.strip()

    if model_type == chatgpt_model_type_memory:
        conversation_system = {"role": "system", "content": ""}
        conversation_user = {"role": "user", "content": input_text}
        if conversation_history == []:
            conversation_history.append(conversation_system)
            conversation_history.append(conversation_user)
        else:
            conversation_history.append(conversation_user)

        response = client.chat.completions.create(
            model="gpt-3.5-turbo-16k",
            temperature=0.25,
            max_tokens=5000,
            frequency_penalty=0.5,
            presence_penalty=0.5,
            top_p=0.88,
            messages=conversation_history,
        )
        response_text = response.choices[0].message.content

        conversation_system = {"role": "system", "content": response_text}
        conversation_history.append(conversation_system)

    if model_type == chatgpt_model_type_gemini:
        if prompt_parts[0] == "":
            prompt_parts[0] += input_text
        else:
            prompt_parts[0] += (" " + input_text)

        response_gemini = generation(prompt_parts)
        response_text = response_gemini.text
        prompt_parts[0] += (" " + response_text)

    return response_text


dall_e_model_type_2 = "Dall-E-2：较快 | 较差"
dall_e_model_type_3 = "Dall-E-3：较慢 | 较好"
def dall_e(input_text, add_text, model_type):
    a_key = "c2stRFo3ZzRadXlzOEhMd2RWZnFYVWJUM0JsYmtGSkFPTm9PNHdPMmJZTTc5Tm5DTVQ5"
    client = OpenAI(api_key=str(base64.b64decode(a_key), encoding="utf-8"))

    complete_text = input_text + "\n" + "增加图片风格的细节或场景的细节：" + add_text
    if model_type == dall_e_model_type_2:
        model_select = "dall-e-2"
    if model_type == dall_e_model_type_3:
        model_select = "dall-e-3"

    response = client.images.generate(
        model=model_select,
        prompt=complete_text,
        size="1024x1024",
        quality="standard",
        n=1,
    )
    image_url = response.data[0].url
    return image_url


def whisper(audio_file):
    a_key = "c2stRFo3ZzRadXlzOEhMd2RWZnFYVWJUM0JsYmtGSkFPTm9PNHdPMmJZTTc5Tm5DTVQ5"
    client = OpenAI(api_key=str(base64.b64decode(a_key), encoding="utf-8"))
    audio_file = open(audio_file, "rb")
    transcript = client.audio.transcriptions.create(model="whisper-1", file=audio_file, response_format="text")

    return transcript


tts_model_type_1 = "tts-1：较快 | 较差"
tts_model_type_1_hd = "tts-1-hd：较慢 | 较好"
tts_gender_male = "男生声音：Alloy"
tts_gender_female = "女生声音：Nova"
def tts(input_text, model_type, gender):
    a_key = "c2stRFo3ZzRadXlzOEhMd2RWZnFYVWJUM0JsYmtGSkFPTm9PNHdPMmJZTTc5Tm5DTVQ5"
    client = OpenAI(api_key=str(base64.b64decode(a_key), encoding="utf-8"))

    timestamp = time.time()
    speech_file_path = "tts_" + str(timestamp) + ".mp3"

    if model_type == "tts-1：较快 | 较差":
        model_select = "tts-1"
    if model_type == "tts-1-hd：较慢 | 较好":
        model_select = "tts-1-hd"
    if gender == "男生声音：Alloy":
        voice_select = "alloy"
    if gender == "女生声音：Nova":
        voice_select = "nova"

    response = client.audio.speech.create(
        model=model_select,
        voice=voice_select,
        input=input_text,
    )
    response.stream_to_file(speech_file_path)
    return speech_file_path, whisper(speech_file_path)


def demo():
    article_thanks = "<br> 🔅Thanks Python. &nbsp;&nbsp;  🔅Thanks OpenAI. &nbsp;&nbsp;  🔅Thanks Gradio. &nbsp;&nbsp;  🔅Thanks Myself. &nbsp;&nbsp;  🔅Thank you ！"
    chat_demo = gradio.Interface(fn=chatgpt,
                                 inputs=[gradio.Textbox(label="💢(必须) ⚡你的对话", lines=3, placeholder="输入内容", show_copy_button=True),
                                         gradio.Radio([chatgpt_model_type_normal, chatgpt_model_type_memory, chatgpt_model_type_gemini],
                                                      label="⚡选择GPT模型", value=chatgpt_model_type_normal),
                                         gradio.Textbox(label="⚡人物角色 / 学术领域（选择 #记忆对话型 生效）", lines=1,
                                                        placeholder="例如：马斯克；人工智能")],
                                 outputs=gradio.Textbox(label="⚡ GPT 对话生成", lines=11,
                                                        show_copy_button=True,
                                                        placeholder="ChatGPT - 是未成熟的AGI智能工具 - 无法保证回答正确 ！仅供参考 😊"),
                                 title="<br>MyDIY_Resume | Gradio：🤖 Chat-with-GPT3.5",
                                 description="✅ 操作指南：在左边的文本框中输入内容，再选择对话的GPT模型类型，如果选择为角色型则可以为GPT设定一种角色或领域作为回答的方向 ！😁",
                                 article=article_thanks)

    image_demo = gradio.Interface(fn=dall_e,
                                  inputs=[gradio.Textbox(label="💢(必须) ⚡描述图片", lines=2, placeholder="输入内容", show_copy_button=True),
                                          gradio.Textbox(label="⚡描述细节", lines=1, placeholder="例如：蔚蓝；科幻", show_copy_button=True),
                                          gradio.Radio([dall_e_model_type_2, dall_e_model_type_3],
                                                       label="⚡选择Dall-E模型", value=dall_e_model_type_2)],
                                  outputs=gradio.Image(label="⚡Dell-E 图片生成"),
                                  title="<br>MyDIY_Resume | Gradio：🤖 Dall-E-Generate-Image",
                                  description="✅ 操作指南：在左边的文本框中输入需要生成图片的内容，再选择生成图片的Dall-E模型，需要注意Dall-E-3模型的请求更容易受到限制 ！😁",
                                  article=article_thanks)

    stt_demo = gradio.Interface(fn=whisper,
                                inputs=[gradio.Audio(label="💢(必须) ⚡上传音频文件", type="filepath")],
                                outputs=gradio.Textbox(label="⚡Whisper 文字生成", lines=10,
                                                       placeholder="Whisper-v1 - 是初代版本模型 - 无法保证转换正确 ！仅供参考 😊",
                                                       show_copy_button=True),
                                title="<br>MyDIY_Resume | Gradio：🤖 Whisper-Speech-to-Text",
                                description="✅ 操作指南：点击左边的音频框进行本地音频文件选取，或者直接将音频文件拖拉到当前音频框，默认的音频转文本模型是Whisper-v1 ！😁",
                                article=article_thanks)

    tts_demo = gradio.Interface(fn=tts,
                                inputs=[gradio.Textbox(label="💢(必须) ⚡你的文本", lines=3, placeholder="输入内容", show_copy_button=True),
                                        gradio.Radio([tts_model_type_1, tts_model_type_1_hd], label="选择TTS模型", value=tts_model_type_1),
                                        gradio.Radio([tts_gender_male, tts_gender_female], label="选择声音性别", value=tts_gender_male)],
                                outputs=[gradio.Audio(label="⚡TTS 音频生成"),
                                         gradio.Textbox(label="生成的音频再经过 Whisper 二次检验", lines=5,
                                                        placeholder="生成的音频再次经过 Whisper 转换用于验证TTS的准确性 ！仅供参考 😊",
                                                        show_copy_button=True)],
                                title="<br>MyDIY_Resume | Gradio：🤖 TTS-Text-to-Speech",
                                description="✅ 操作指南：在左边的文本框中输入需要转音频的内容，再选择生成音频的TTS模型，两个模型实际差异不大，最后为模型选择输出男声还是女声 ！😁",
                                article=article_thanks)

    GUI = gradio.TabbedInterface([chat_demo, image_demo, stt_demo, tts_demo],
                                 ["1️⃣ Chat-with-GPT：对话学习", "2️⃣ Generate-Image：生成图片", "3️⃣ Speech-to-Text：语音转文本", "4️⃣ Text-to-Speech：文本转语音"],
                                 css="footer {visibility: hidden}")

    GUI.launch(debug=True, share=False)


if __name__ == "__main__":
    demo()