Upload 3 files
Browse files- app.py +217 -0
- config.py +39 -0
- requirements.txt +8 -0
app.py
ADDED
|
@@ -0,0 +1,217 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
|
| 3 |
+
import torch
|
| 4 |
+
import requests
|
| 5 |
+
import os
|
| 6 |
+
|
| 7 |
+
# ============================================
|
| 8 |
+
# MODEL SETUP - BharatGen Param-1 Instruct
|
| 9 |
+
# ============================================
|
| 10 |
+
|
| 11 |
+
print("Loading BharatGen Param-1...")
|
| 12 |
+
|
| 13 |
+
# Use Instruct version for better responses
|
| 14 |
+
PARAM1_MODEL = "bharatgenai/Param-1-2.9B-Instruct"
|
| 15 |
+
HF_TOKEN = os.getenv("HF_TOKEN")
|
| 16 |
+
|
| 17 |
+
# Load tokenizer and model
|
| 18 |
+
tokenizer = AutoTokenizer.from_pretrained(PARAM1_MODEL, trust_remote_code=True)
|
| 19 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 20 |
+
PARAM1_MODEL,
|
| 21 |
+
trust_remote_code=True,
|
| 22 |
+
torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
|
| 23 |
+
device_map="auto",
|
| 24 |
+
low_cpu_mem_usage=True
|
| 25 |
+
)
|
| 26 |
+
|
| 27 |
+
# STT Model - AI4Bharat IndicConformer
|
| 28 |
+
STT_API = "https://api-inference.huggingface.co/models/ai4bharat/conformer_multilingual"
|
| 29 |
+
|
| 30 |
+
# ============================================
|
| 31 |
+
# SYSTEM PROMPT FOR WOMEN EMPOWERMENT
|
| 32 |
+
# ============================================
|
| 33 |
+
|
| 34 |
+
SYSTEM_PROMPT = """आप एक सहायक AI हैं जो भारतीय महिलाओं की सुरक्षा और सशक्तिकरण के लिए बनाई गई है।
|
| 35 |
+
|
| 36 |
+
निम्नलिखित विश्लेषण प्रदान करें:
|
| 37 |
+
1. भावना श्रेणी (Emergency/Distress/Help Needed/Support/Positive)
|
| 38 |
+
2. सहानुभूतिपूर्ण प्रतिक्रिया (1-2 वाक्य)
|
| 39 |
+
3. 2-3 ठोस कार्रवाई
|
| 40 |
+
4. प्रासंगिक हेल्पलाइन (यदि आवश्यक हो)
|
| 41 |
+
|
| 42 |
+
महत्वपूर्ण हेल्पलाइन:
|
| 43 |
+
- महिला हेल्पलाइन: 181
|
| 44 |
+
- पुलिस आपातकाल: 100, 112
|
| 45 |
+
- मानसिक स्वास्थ्य: NIMHANS 080-46110007
|
| 46 |
+
|
| 47 |
+
उत्तर संक्षिप्त, व्यावहारिक और सहायक हो।"""
|
| 48 |
+
|
| 49 |
+
# ============================================
|
| 50 |
+
# CORE FUNCTIONS
|
| 51 |
+
# ============================================
|
| 52 |
+
|
| 53 |
+
def transcribe_audio(audio_path):
|
| 54 |
+
"""Convert speech to text using IndicConformer"""
|
| 55 |
+
try:
|
| 56 |
+
with open(audio_path, "rb") as f:
|
| 57 |
+
data = f.read()
|
| 58 |
+
|
| 59 |
+
response = requests.post(
|
| 60 |
+
STT_API,
|
| 61 |
+
headers={"Authorization": f"Bearer {HF_TOKEN}"},
|
| 62 |
+
data=data,
|
| 63 |
+
timeout=30
|
| 64 |
+
)
|
| 65 |
+
|
| 66 |
+
result = response.json()
|
| 67 |
+
|
| 68 |
+
# Handle different response formats
|
| 69 |
+
if isinstance(result, dict) and "text" in result:
|
| 70 |
+
return result["text"]
|
| 71 |
+
elif isinstance(result, list) and len(result) > 0:
|
| 72 |
+
return result[0].get("text", "")
|
| 73 |
+
else:
|
| 74 |
+
return ""
|
| 75 |
+
|
| 76 |
+
except Exception as e:
|
| 77 |
+
print(f"STT Error: {e}")
|
| 78 |
+
return ""
|
| 79 |
+
|
| 80 |
+
def analyze_sentiment_and_suggest(text):
|
| 81 |
+
"""Use BharatGen Param-1 for sentiment analysis and suggestions"""
|
| 82 |
+
try:
|
| 83 |
+
# Create prompt for BharatGen
|
| 84 |
+
prompt = f"""{SYSTEM_PROMPT}
|
| 85 |
+
|
| 86 |
+
महिला का कथन: {text}
|
| 87 |
+
|
| 88 |
+
विश्लेषण और सुझाव:"""
|
| 89 |
+
|
| 90 |
+
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
|
| 91 |
+
|
| 92 |
+
with torch.no_grad():
|
| 93 |
+
outputs = model.generate(
|
| 94 |
+
**inputs,
|
| 95 |
+
max_new_tokens=250,
|
| 96 |
+
do_sample=True,
|
| 97 |
+
top_k=50,
|
| 98 |
+
top_p=0.95,
|
| 99 |
+
temperature=0.7,
|
| 100 |
+
eos_token_id=tokenizer.eos_token_id,
|
| 101 |
+
pad_token_id=tokenizer.pad_token_id
|
| 102 |
+
)
|
| 103 |
+
|
| 104 |
+
generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 105 |
+
|
| 106 |
+
# Extract only the response after the prompt
|
| 107 |
+
response = generated_text.split("विश्लेषण और सुझाव:")[-1].strip()
|
| 108 |
+
|
| 109 |
+
return response
|
| 110 |
+
|
| 111 |
+
except Exception as e:
|
| 112 |
+
print(f"Analysis Error: {e}")
|
| 113 |
+
return f"⚠️ विश्लेषण में त्रुटि। कृपया पुनः प्रयास करें।\nError: {str(e)}"
|
| 114 |
+
|
| 115 |
+
def process_voice_input(audio):
|
| 116 |
+
"""Main pipeline: Voice → Text → Analysis → Suggestions"""
|
| 117 |
+
|
| 118 |
+
if audio is None:
|
| 119 |
+
return "⚠️ कृपया अपनी आवाज़ रिकॉर्ड करें।\n⚠️ Please record your voice."
|
| 120 |
+
|
| 121 |
+
try:
|
| 122 |
+
# Step 1: Speech to Text
|
| 123 |
+
print("Converting speech to text...")
|
| 124 |
+
transcribed_text = transcribe_audio(audio)
|
| 125 |
+
|
| 126 |
+
if not transcribed_text or len(transcribed_text.strip()) < 3:
|
| 127 |
+
return """⚠️ आपकी आवाज़ स्पष्ट नहीं सुनाई दी। कृपया फिर से बोलें।
|
| 128 |
+
⚠️ Could not understand your voice clearly. Please speak again."""
|
| 129 |
+
|
| 130 |
+
# Step 2: Sentiment Analysis and Suggestions
|
| 131 |
+
print(f"Analyzing: {transcribed_text}")
|
| 132 |
+
suggestions = analyze_sentiment_and_suggest(transcribed_text)
|
| 133 |
+
|
| 134 |
+
# Format output
|
| 135 |
+
output = f"""📝 **आपने कहा / What you said:**
|
| 136 |
+
{transcribed_text}
|
| 137 |
+
|
| 138 |
+
━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
| 139 |
+
|
| 140 |
+
🤖 **विश्लेषण और सुझाव / Analysis & Suggestions:**
|
| 141 |
+
{suggestions}
|
| 142 |
+
|
| 143 |
+
━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
| 144 |
+
|
| 145 |
+
📞 **आपातकालीन संपर्क / Emergency Contacts:**
|
| 146 |
+
• महिला हेल्पलाइन: 181
|
| 147 |
+
• पुलिस आपातकाल: 100, 112
|
| 148 |
+
• मानसिक स्वास्थ्य: 080-46110007
|
| 149 |
+
"""
|
| 150 |
+
|
| 151 |
+
return output
|
| 152 |
+
|
| 153 |
+
except Exception as e:
|
| 154 |
+
return f"""❌ त्रुटि / Error: {str(e)}
|
| 155 |
+
|
| 156 |
+
कृपया पुनः प्रयास करें या सीधे 181 पर कॉल करें।
|
| 157 |
+
Please try again or call 181 directly for immediate help."""
|
| 158 |
+
|
| 159 |
+
# ============================================
|
| 160 |
+
# GRADIO INTERFACE
|
| 161 |
+
# ============================================
|
| 162 |
+
|
| 163 |
+
# Custom CSS for better UI
|
| 164 |
+
custom_css = """
|
| 165 |
+
.gradio-container {
|
| 166 |
+
font-family: 'Inter', sans-serif;
|
| 167 |
+
}
|
| 168 |
+
.output-box {
|
| 169 |
+
font-size: 16px;
|
| 170 |
+
line-height: 1.6;
|
| 171 |
+
}
|
| 172 |
+
"""
|
| 173 |
+
|
| 174 |
+
# Create interface
|
| 175 |
+
demo = gr.Interface(
|
| 176 |
+
fn=process_voice_input,
|
| 177 |
+
inputs=gr.Audio(
|
| 178 |
+
sources=["microphone"],
|
| 179 |
+
type="filepath",
|
| 180 |
+
label="🎤 अपनी आवाज़ रिकॉर्ड करें / Record Your Voice",
|
| 181 |
+
show_label=True
|
| 182 |
+
),
|
| 183 |
+
outputs=gr.Textbox(
|
| 184 |
+
label="📋 विश्लेषण और सुझाव / Analysis & Suggestions",
|
| 185 |
+
lines=20,
|
| 186 |
+
show_label=True,
|
| 187 |
+
elem_classes="output-box"
|
| 188 |
+
),
|
| 189 |
+
title="🌸 महिला सशक्तिकरण सहायक / Women Empowerment Assistant",
|
| 190 |
+
description="""
|
| 191 |
+
**हिंदी और अंग्रेजी में बोलें | Speak in Hindi or English**
|
| 192 |
+
|
| 193 |
+
यह एक AI-आधारित सहायक है जो महिलाओं की चिंताओं को सुनता है और व्यावहारिक सुझाव देता है।
|
| 194 |
+
|
| 195 |
+
*This AI assistant listens to women's concerns and provides actionable suggestions.*
|
| 196 |
+
|
| 197 |
+
🔒 **गोपनीयता:** आपकी आवाज़ सुरक्षित है। | Your voice is secure.
|
| 198 |
+
""",
|
| 199 |
+
theme=gr.themes.Soft(
|
| 200 |
+
primary_hue="pink",
|
| 201 |
+
secondary_hue="purple"
|
| 202 |
+
),
|
| 203 |
+
examples=[],
|
| 204 |
+
css=custom_css,
|
| 205 |
+
allow_flagging="never"
|
| 206 |
+
)
|
| 207 |
+
|
| 208 |
+
# ============================================
|
| 209 |
+
# LAUNCH
|
| 210 |
+
# ============================================
|
| 211 |
+
|
| 212 |
+
if __name__ == "__main__":
|
| 213 |
+
demo.launch(
|
| 214 |
+
share=False,
|
| 215 |
+
server_name="0.0.0.0",
|
| 216 |
+
server_port=7860
|
| 217 |
+
)
|
config.py
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Configuration file for Women Empowerment Voice Assistant
|
| 3 |
+
"""
|
| 4 |
+
|
| 5 |
+
# Model Configuration
|
| 6 |
+
BHARATGEN_MODEL = "bharatgenai/Param-1-2.9B-Instruct"
|
| 7 |
+
STT_MODEL = "ai4bharat/conformer_multilingual"
|
| 8 |
+
|
| 9 |
+
# Supported Languages (current)
|
| 10 |
+
SUPPORTED_LANGUAGES = [
|
| 11 |
+
"Hindi (हिंदी)",
|
| 12 |
+
"English (अंग्रेजी)",
|
| 13 |
+
"Marathi (मराठी)",
|
| 14 |
+
"Tamil (தமிழ்)",
|
| 15 |
+
"Malayalam (മലയാളം)",
|
| 16 |
+
"Bengali (বাংলা)",
|
| 17 |
+
"Punjabi (ਪੰਜਾਬੀ)",
|
| 18 |
+
"Gujarati (ગુજરાતી)",
|
| 19 |
+
"Telugu (తెలుగు)",
|
| 20 |
+
"Kannada (ಕನ್ನಡ)"
|
| 21 |
+
]
|
| 22 |
+
|
| 23 |
+
# Emergency Contacts
|
| 24 |
+
HELPLINES = {
|
| 25 |
+
"women_helpline": "181",
|
| 26 |
+
"police_emergency": "100, 112",
|
| 27 |
+
"child_helpline": "1098",
|
| 28 |
+
"mental_health": "080-46110007",
|
| 29 |
+
"domestic_violence": "181"
|
| 30 |
+
}
|
| 31 |
+
|
| 32 |
+
# Generation Parameters
|
| 33 |
+
GEN_CONFIG = {
|
| 34 |
+
"max_new_tokens": 250,
|
| 35 |
+
"temperature": 0.7,
|
| 36 |
+
"top_k": 50,
|
| 37 |
+
"top_p": 0.95,
|
| 38 |
+
"do_sample": True
|
| 39 |
+
}
|
requirements.txt
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
gradio==4.44.0
|
| 2 |
+
transformers==4.45.0
|
| 3 |
+
torch==2.4.0
|
| 4 |
+
accelerate==0.34.0
|
| 5 |
+
requests==2.32.0
|
| 6 |
+
soundfile==0.12.1
|
| 7 |
+
librosa==0.10.1
|
| 8 |
+
sentencepiece==0.2.0
|