File size: 9,901 Bytes
3bddc3f
bc91045
3bddc3f
4b286f3
94421aa
3bddc3f
bc91045
3bddc3f
6acf91d
1ba22b4
3bddc3f
1ba22b4
4b286f3
 
 
 
3bddc3f
 
 
 
 
 
 
bc91045
 
 
3bddc3f
 
 
bc91045
 
 
3bddc3f
1ba22b4
6acf91d
 
 
aa07a0a
6acf91d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4b286f3
bc91045
 
1ba22b4
bc91045
 
 
 
1ba22b4
3bddc3f
bc91045
 
1ba22b4
3bddc3f
 
 
 
bc91045
4b286f3
 
 
bc91045
3bddc3f
45e1616
bc91045
 
1ba22b4
6acf91d
3bddc3f
bc91045
 
39cf431
 
 
 
027e874
94421aa
027e874
 
bc91045
027e874
3bddc3f
 
82502f1
 
149da6f
 
 
82502f1
149da6f
 
 
 
94421aa
149da6f
82502f1
149da6f
82502f1
149da6f
94421aa
82502f1
94421aa
1ba22b4
82502f1
149da6f
 
 
 
 
 
 
 
 
 
82502f1
149da6f
82502f1
149da6f
 
 
 
 
82502f1
 
149da6f
 
 
82502f1
149da6f
 
 
 
94421aa
149da6f
82502f1
149da6f
82502f1
149da6f
94421aa
82502f1
94421aa
149da6f
82502f1
1ba22b4
 
149da6f
 
 
 
 
 
 
 
82502f1
149da6f
82502f1
149da6f
 
 
 
 
82502f1
1ba22b4
 
82502f1
 
027e874
bc91045
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
import logging
import os
import re
import warnings
from pathlib import Path

import gradio as gr
import requests
import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline, Pipeline

from templates import starting_app_code, update_iframe_js, copy_snippet_js, download_code_js, load_js, DemoType

# Filter the UserWarning raised by the audio component.
warnings.filterwarnings("ignore", message='Trying to convert audio automatically from int32 to 16-bit int format')

logging.basicConfig(
    level=logging.INFO,  # Set the logging level to INFO or any other desired level
    format="%(asctime)s - %(message)s",  # Define the log message format
    datefmt="%Y-%m-%d %H:%M:%S",  # Define the timestamp format
)

logger = logging.getLogger("my_logger")

HF_TOKEN = os.getenv("HF_TOKEN")

if not HF_TOKEN:
    raise Exception("HF_TOKEN environment variable is required to call remote API.")

API_URL = "https://api-inference.huggingface.co/models/HuggingFaceH4/zephyr-7b-beta"
headers = {"Authorization": f"Bearer {HF_TOKEN}"}


def init_speech_to_text_model() -> Pipeline:
    device = "cuda:0" if torch.cuda.is_available() else "cpu"
    torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

    model_id = "distil-whisper/distil-medium.en"
    model = AutoModelForSpeechSeq2Seq.from_pretrained(
        model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
    )
    model.to(device)
    processor = AutoProcessor.from_pretrained(model_id)
    return pipeline(
        "automatic-speech-recognition",
        model=model,
        tokenizer=processor.tokenizer,
        feature_extractor=processor.feature_extractor,
        max_new_tokens=128,
        torch_dtype=torch_dtype,
        device=device,
    )


whisper_pipe = init_speech_to_text_model()

code_pattern = re.compile(r'```python\n(.*?)```', re.DOTALL)


def query(payload: dict):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()


def generate_text(code: str, prompt: str) -> (str, str, str):
    logger.info(f"Calling API with prompt:\n{prompt}")
    prompt = f"```python\n{code}```\nGiven the code above return only updated code for the following request:\n{prompt}\n<|assistant|>"
    params = {"max_new_tokens": 512}
    output = query({"inputs": prompt, "parameters": params})
    if 'error' in output:
        logger.warning(f'Language model call failed: {output["error"]}')
        raise gr.Warning(f'Language model call failed: {output["error"]}')
    logger.info(f'API RESPONSE\n{output[0]["generated_text"]}')
    assistant_reply = output[0]["generated_text"].split('<|assistant|>')[1]
    match = re.search(code_pattern, assistant_reply)
    if not match:
        return assistant_reply, code, None
    new_code = match.group(1)
    logger.info(f'NEW CODE:\nnew_code')
    return assistant_reply, new_code, None


def transcribe(audio: str) -> (str, str):
    result = whisper_pipe(audio)
    return result["text"], None


def copy_notify(code):
    gr.Info("App code snippet copied!")


def add_hotkeys() -> str:
    return Path("hotkeys.js").read_text()


with gr.Blocks() as demo:
    gr.Markdown("<h1 id=\"TEST\" align=\"center\">KiteWind πŸͺπŸƒ</h1>")
    gr.Markdown(
        "<h4 align=\"center\">Chat-assisted web app creator by <a href=\"https://huggingface.co/gstaff\">@gstaff</a></h4>")
    selectedTab = gr.State(value='gradio-lite')
    with gr.Tab('Gradio (gradio-lite)') as gradio_lite_tab:
        with gr.Row():
            with gr.Column():
                gr.Markdown("## 1. Run your app in the browser!")
                gr.HTML(value='<div id="gradioDemoDiv"></div>')
        gr.Markdown("## 2. Customize using voice requests!")
        with gr.Row():
            with gr.Column():
                with gr.Group():
                    in_audio = gr.Audio(label="Record a voice request (click or press ctrl + ` to start/stop)", source='microphone', type='filepath', elem_classes=["record-btn"])
                    in_prompt = gr.Textbox(label="Or type a text request and press Enter",
                                           placeholder="Need an idea? Try one of these:\n- Add a button to reverse the name\n- Change the greeting to Spanish\n- Put the reversed name output into a separate textbox")
                out_text = gr.TextArea(label="Chat Assistant Response")
                clear = gr.ClearButton([in_prompt, in_audio, out_text])
            with gr.Column():
                code_area = gr.Code(label="App Code - You can also edit directly and then click Update App or ctrl + space",
                                    language='python', value=starting_app_code(DemoType.GRADIO))
                update_btn = gr.Button("Update App (Ctrl + Space)", variant="primary", elem_classes=["update-btn"])
                code_update_params = {'fn': None, 'inputs': code_area, 'outputs': None,
                                      '_js': update_iframe_js(DemoType.GRADIO)}
                gen_text_params = {'fn': generate_text, 'inputs': [code_area, in_prompt],
                                   'outputs': [out_text, code_area]}
                transcribe_params = {'fn': transcribe, 'inputs': [in_audio], 'outputs': [in_prompt, in_audio]}
                update_btn.click(**code_update_params)
                in_prompt.submit(**gen_text_params).then(**code_update_params)
                in_audio.stop_recording(**transcribe_params).then(**gen_text_params).then(**code_update_params)
        with gr.Row():
            with gr.Column():
                gr.Markdown("## 3. Export your app to share!")
                copy_snippet_btn = gr.Button("Copy app snippet to paste in another page")
                copy_snippet_btn.click(copy_notify, code_area, None, _js=copy_snippet_js(DemoType.GRADIO))
                download_btn = gr.Button("Download app as a standalone file")
                download_btn.click(None, code_area, None, _js=download_code_js(DemoType.GRADIO))
        with gr.Row():
            with gr.Column():
                gr.Markdown("## Current limitations")
                with gr.Accordion("Click to view", open=False):
                    gr.Markdown(
                        "- Only gradio-lite apps using the python standard libraries and gradio are supported\n- The chat hasn't been tuned on gradio library data; it may make mistakes\n- The app needs to fully reload each time it is changed")
    with gr.Tab('Streamlit (stlite)') as stlite_tab:
        with gr.Row():
            with gr.Column():
                gr.Markdown("## 1. Run your app in the browser!")
                gr.HTML(value='<div id="stliteDemoDiv"></div>')
        gr.Markdown("## 2. Customize using voice requests!")
        with gr.Row():
            with gr.Column():
                with gr.Group():
                    in_audio = gr.Audio(label="Record a voice request (click or press ctrl + ` to start/stop)", source='microphone', type='filepath', elem_classes=["record-btn"])
                    in_prompt = gr.Textbox(label="Or type a text request and press Enter",
                                           placeholder="Need an idea? Try one of these:\n- Add a button to reverse the name\n- Change the greeting to Spanish\n- Make the button primary")
                out_text = gr.TextArea(label="Chat Assistant Response")
                clear_btn = gr.ClearButton([in_prompt, in_audio, out_text])
            with gr.Column():
                code_area = gr.Code(label="App Code - You can also edit directly and then click Update App or ctrl + space",
                                    language='python', value=starting_app_code(DemoType.STREAMLIT))
                update_btn = gr.Button("Update App (Ctrl + Space)", variant="primary", elem_classes=["update-btn"])
                code_update_params = {'fn': None, 'inputs': code_area, 'outputs': None,
                                      '_js': update_iframe_js(DemoType.STREAMLIT)}
                gen_text_params = {'fn': generate_text, 'inputs': [code_area, in_prompt],
                                   'outputs': [out_text, code_area]}
                transcribe_params = {'fn': transcribe, 'inputs': [in_audio], 'outputs': [in_prompt, in_audio]}
                update_btn.click(**code_update_params)
                in_prompt.submit(**gen_text_params).then(**code_update_params)
                in_audio.stop_recording(**transcribe_params).then(**gen_text_params).then(**code_update_params)
        with gr.Row():
            with gr.Column():
                gr.Markdown("## 3. Export your app to share!")
                copy_snippet_btn = gr.Button("Copy app snippet to paste in another page")
                copy_snippet_btn.click(copy_notify, code_area, None, _js=copy_snippet_js(DemoType.STREAMLIT))
                download_btn = gr.Button("Download app as a standalone file")
                download_btn.click(None, code_area, None, _js=download_code_js(DemoType.STREAMLIT))
        with gr.Row():
            with gr.Column():
                gr.Markdown("## Current limitations")
                with gr.Accordion("Click to view", open=False):
                    gr.Markdown(
                        "- Only Streamlit apps using libraries available in pyodide are supported\n- The chat hasn't been tuned on Streamlit library data; it may make mistakes")
    gradio_lite_tab.select(lambda: "gradio-lite", None, selectedTab).then(None, None, None,
                                                                          _js=load_js(DemoType.GRADIO))
    stlite_tab.select(lambda: "stlite", None, selectedTab).then(None, None, None, _js=load_js(DemoType.STREAMLIT))
    demo.load(None, None, None, _js=load_js(DemoType.GRADIO))
    demo.load(None, None, None, _js=add_hotkeys())
    demo.css = "footer {visibility: hidden}"

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