Spaces:
Sleeping
Sleeping
gradio file
Browse files
app.py
ADDED
@@ -0,0 +1,130 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import faster_whisper
|
3 |
+
import gradio as gr
|
4 |
+
from dotenv import load_dotenv
|
5 |
+
from huggingface_hub import InferenceClient
|
6 |
+
|
7 |
+
# Load API key dari .env
|
8 |
+
load_dotenv()
|
9 |
+
HF_API_KEY = os.getenv("HF_API_KEY")
|
10 |
+
|
11 |
+
if not HF_API_KEY:
|
12 |
+
raise ValueError("API Key Hugging Face tidak ditemukan. Pastikan file .env berisi HF_API_KEY.")
|
13 |
+
|
14 |
+
# Inisialisasi klien API Hugging Face
|
15 |
+
huggingface_client = InferenceClient(api_key=HF_API_KEY)
|
16 |
+
|
17 |
+
# Load Faster Whisper model versi large
|
18 |
+
model = faster_whisper.WhisperModel("turbo", device="cpu", compute_type="int8")
|
19 |
+
|
20 |
+
def save_to_file(content, filename):
|
21 |
+
with open(filename, 'w', encoding='utf-8') as file:
|
22 |
+
file.write(content)
|
23 |
+
return filename
|
24 |
+
|
25 |
+
def transcribe_audio(audio_path):
|
26 |
+
"""Transkripsi audio menggunakan Faster Whisper tanpa koreksi model Hugging Face."""
|
27 |
+
segments, _ = model.transcribe(audio_path)
|
28 |
+
raw_transcription = " ".join(segment.text for segment in segments)
|
29 |
+
return raw_transcription, save_to_file(raw_transcription, 'transcription_large.txt'), audio_path
|
30 |
+
|
31 |
+
def generate_soap_summary(transcription_text):
|
32 |
+
"""Membuat ringkasan SOAP dari teks transkripsi."""
|
33 |
+
template = """
|
34 |
+
Anda adalah asisten medis yang membantu dokter dalam menyusun catatan SOAP berdasarkan percakapan dokter dan pasien.
|
35 |
+
Ringkaskan dalam bentuk paragraf tanpa adanya bullet point dan gunakan bahasa Indonesia.
|
36 |
+
Harap buat ringkasan dalam format berikut:
|
37 |
+
|
38 |
+
Subjective:
|
39 |
+
Objective:
|
40 |
+
Assessment:
|
41 |
+
Plan:
|
42 |
+
|
43 |
+
### Percakapan:
|
44 |
+
{dialogue}
|
45 |
+
|
46 |
+
### Catatan SOAP:
|
47 |
+
"""
|
48 |
+
messages = [{"role": "user", "content": template.format(dialogue=transcription_text)}]
|
49 |
+
response = huggingface_client.chat.completions.create(
|
50 |
+
model="mistralai/Mixtral-8x7B-Instruct-v0.1",
|
51 |
+
messages=messages,
|
52 |
+
max_tokens=1000,
|
53 |
+
stream=False
|
54 |
+
)
|
55 |
+
soap = response.choices[0].message.content.strip()
|
56 |
+
return soap, save_to_file(soap, 'soap_summary.txt')
|
57 |
+
|
58 |
+
def detect_medical_tags(transcription_text):
|
59 |
+
"""Mendeteksi tags Diagnosis, Obat, Hasil Lab, dan Radiologi."""
|
60 |
+
template = """
|
61 |
+
Identifikasi dan berikan luaran dalam bahasa indonesia tags berikut dari percakapan:
|
62 |
+
Diagnosis:
|
63 |
+
Obat:
|
64 |
+
Hasil Lab:
|
65 |
+
Radiologi:
|
66 |
+
|
67 |
+
### Percakapan:
|
68 |
+
{dialogue}
|
69 |
+
"""
|
70 |
+
messages = [{"role": "user", "content": template.format(dialogue=transcription_text)}]
|
71 |
+
response = huggingface_client.chat.completions.create(
|
72 |
+
model="mistralai/Mixtral-8x7B-Instruct-v0.1",
|
73 |
+
messages=messages,
|
74 |
+
max_tokens=500,
|
75 |
+
stream=False
|
76 |
+
)
|
77 |
+
tags = response.choices[0].message.content.strip()
|
78 |
+
return tags, save_to_file(tags, 'medical_tags.txt')
|
79 |
+
|
80 |
+
# Antarmuka Gradio
|
81 |
+
with gr.Blocks(title="AI-based Medical SOAP Summarization and Tag Detection with Faster Whisper Large") as app:
|
82 |
+
gr.Markdown("## Medical SOAP Summarization and Tag Detection with Faster Whisper Large")
|
83 |
+
|
84 |
+
with gr.Row():
|
85 |
+
with gr.Column():
|
86 |
+
audio_input = gr.Audio("microphone", type="filepath", label="🎙️ Rekam Suara")
|
87 |
+
transcribe_button = gr.Button("🎧 Transkripsi dengan Whisper Large")
|
88 |
+
transcription_edit_box = gr.Textbox(label="📄 Hasil Transkripsi (Faster Whisper Large) - Bisa Diedit", lines=12, interactive=True)
|
89 |
+
update_transcription_button = gr.Button("💾 Simpan Hasil Edit")
|
90 |
+
soap_button = gr.Button("📝 Buat SOAP")
|
91 |
+
tags_button = gr.Button("🏷️ Deteksi Tags")
|
92 |
+
|
93 |
+
with gr.Column():
|
94 |
+
soap_output = gr.Textbox(label="📃 Hasil SOAP", lines=10, interactive=False)
|
95 |
+
tags_output = gr.Textbox(label="🏷️ Hasil Tags Diagnosis, Obat, Hasil Lab, Radiologi", lines=10, interactive=False)
|
96 |
+
download_audio = gr.File(label="⬇️ Download Rekaman")
|
97 |
+
download_transcription = gr.File(label="⬇️ Download Transkripsi")
|
98 |
+
download_soap = gr.File(label="⬇️ Download SOAP")
|
99 |
+
download_tags = gr.File(label="⬇️ Download Tags")
|
100 |
+
|
101 |
+
# Tombol Transkripsi
|
102 |
+
transcribe_button.click(
|
103 |
+
transcribe_audio,
|
104 |
+
inputs=[audio_input],
|
105 |
+
outputs=[transcription_edit_box, download_transcription, download_audio]
|
106 |
+
)
|
107 |
+
|
108 |
+
# Tombol Simpan Hasil Edit
|
109 |
+
update_transcription_button.click(
|
110 |
+
lambda text: (text, save_to_file(text, 'user_edited_transcription.txt')),
|
111 |
+
inputs=[transcription_edit_box],
|
112 |
+
outputs=[transcription_edit_box, download_transcription]
|
113 |
+
)
|
114 |
+
|
115 |
+
# Tombol SOAP
|
116 |
+
soap_button.click(
|
117 |
+
generate_soap_summary,
|
118 |
+
inputs=[transcription_edit_box],
|
119 |
+
outputs=[soap_output, download_soap]
|
120 |
+
)
|
121 |
+
|
122 |
+
# Tombol Tags
|
123 |
+
tags_button.click(
|
124 |
+
detect_medical_tags,
|
125 |
+
inputs=[transcription_edit_box],
|
126 |
+
outputs=[tags_output, download_tags]
|
127 |
+
)
|
128 |
+
|
129 |
+
# Jalankan aplikasi
|
130 |
+
app.launch()
|