app.py

import gradio as gr
import httpx


async def call_api(text: str, audio_path: str):
    # 读取音频文件
    with open(audio_path, "rb") as f:
        audio_bytes = f.read()

    # 发送到后端API
    async with httpx.AsyncClient() as client:
        files = {"audio_file": (audio_path, audio_bytes)}
        data = {"text": text}
        response = await client.post(
            "http://36.151.70.8:30113/process/",
            files=files,
            data=data
        )

    return response.json()["result"]


def load_examples():
    return [
        ["Can you turn my English into German?", "./show_case/common_voice_en_19664034.mp3"],  # En-De
        ["Can you identify the initial word that connects to 'currency_name' in this audio clip?",
         "./show_case/audio-1434542201-headset.wav"],  # ER
        ["What do you think the speaker's message is intended to be in this audio?",
         "./show_case/audio-1434542201-headset.wav"],  # IC
        ["What does the person say?", "./show_case/p225_002.wav"],  # DFake
        ["Assess whether this speech's pronunciation is Real or Fake.", "./show_case/Fake.wav"],  # DFake
        [
            "What emotional weight does the speaker's tone carry?\nPick one answer from A, B, C, and D.\nA: fear\nB: sadness\nC: joy\nD: neutral",
            "./show_case/SER(emotion)_example.wav"],  # SER(emotion)
        [
            "Choose the most suitable answer from options A, B, C, and D to respond the question in next line, you may only choose A or B or C or D.\nThe number of speakers delivering this speech is what?\nA. 4\nB. 2\nC.1\nD. 3",
            "./show_case/SNV_example.wav"],  # SNV
        ["Identify the language of the conversation you just heard.", "./show_case/SLR_example.wav"],  # SLR
        ["tell the gender of the speaker in this audio.", "./show_case/SGR_018.wav"],  # SGR
        ["What's the sound we're hearing in this audio from?", "./show_case/Sound_Vocal_example.wav"],  # Sound_vocal
        ["What is your best guess at the setting of this sound clip?", "./show_case/Scene_example.wav"],  # Sound_cochl
        [
            "Choose the most suitable answer from options A, B, C, and D to respond the question in next line, Please think step by step and you may only choose A or B or C or D.\nRecognize the segment where 'project' is spoken by the speaker.\nA. [5.28, 5.39]\nB. [0.92, 1.39]\nC. [4.75, 5.28]\nD. [3.86, 4.23]",
            "./show_case/SG_audio_1.wav"],  # SG
        ["What type of business does the first person's son have?", "./show_case/SFT_Fisher_example.wav"]  # SFT_Fisher
    ]


iface = gr.Interface(
    fn=call_api,
    inputs=[
        gr.Textbox(label="Enter text instruction", value="What does the person say?"),
        gr.Audio(type="filepath", label="Upload Audio", value="./show_case/p225_002.wav")
    ],
    outputs=gr.Textbox(label="Model output"),
    examples=[],  # Initially no examples shown
    allow_flagging="never"
)

# Add a button to load examples
iface.add_button("Show Example", load_examples)
iface.launch()

if __name__ == '__main__':
    pass