Moved vocca_ai to root for Hugging Face deployment

app.py

@@ -4,7 +4,9 @@ from vocca_ai.sentiment import analyze_sentiment
 from vocca_ai.db_handler import log_call, fetch_recent_calls
 import streamlit as st
 from vocca_ai.preprocess import priority_score
-from vocca_ai.intent_classifier import classify_intent  # Now uses MiniLM
+from vocca_ai.intent_classifier import classify_intent 
+
+
 
 import sys
 import os

vocca_ai/ai_response.py

@@ -0,0 +1,21 @@
+from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
+
+# Load a high-quality summarization model
+model_name = "google/flan-t5-large"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
+
+def generate_call_summary(transcript):
+    """
+    Generates a structured and useful summary of the call.
+    """
+    input_text = f"Summarize this medical call conversation:\n{transcript}"
+    inputs = tokenizer(input_text, return_tensors="pt", truncation=True)
+    outputs = model.generate(**inputs, max_length=100, min_length=20, length_penalty=2.0, num_beams=5)
+
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+# Example Usage
+if __name__ == "__main__":
+    sample_text = "Patient: Hi, I need to schedule an appointment as soon as possible. I’ve been feeling really weak and dizzy for the past few days."
+    print(f"Call Summary: {generate_call_summary(sample_text)}")

vocca_ai/db_handler.py

@@ -0,0 +1,21 @@
+import sqlite3
+
+# Allow database access from multiple threads
+conn = sqlite3.connect("call_logs.db", check_same_thread=False)
+c = conn.cursor()
+
+# Create Table for Logging Calls
+c.execute('''CREATE TABLE IF NOT EXISTS call_logs
+             (id INTEGER PRIMARY KEY, transcript TEXT, intent TEXT, priority TEXT, sentiment TEXT, ai_response TEXT)''')
+conn.commit()
+
+def log_call(transcript, intent, priority, sentiment, ai_response):
+    """Save the call analysis results to database."""
+    c.execute("INSERT INTO call_logs (transcript, intent, priority, sentiment, ai_response) VALUES (?, ?, ?, ?, ?)",
+              (transcript, intent, priority, sentiment, ai_response))
+    conn.commit()
+
+def fetch_recent_calls(limit=5):
+    """Fetch latest logged calls from database."""
+    c.execute("SELECT * FROM call_logs ORDER BY id DESC LIMIT ?", (limit,))
+    return c.fetchall()

vocca_ai/intent_classifier.py

@@ -0,0 +1,23 @@
+def classify_intent(text):
+    """
+    Classifies the intent based on keyword matching (simple rule-based approach).
+    """
+    text_lower = text.lower()
+
+    if any(word in text_lower for word in ["urgent", "emergency", "immediate", "asap", "critical"]):
+        return "urgent medical need"
+    if any(word in text_lower for word in ["appointment", "schedule", "book a visit", "doctor visit"]):
+        return "non-urgent appointment"
+    if any(word in text_lower for word in ["billing", "charge", "invoice", "cost"]):
+        return "billing inquiry"
+    if any(word in text_lower for word in ["insurance", "coverage", "policy"]):
+        return "insurance information"
+    if any(word in text_lower for word in ["pain", "health issue", "fever", "symptom", "sick"]):
+        return "medical advice"
+
+    return "general inquiry"
+
+# Example Usage
+if __name__ == "__main__":
+    sample_text = "I need to book an appointment for my uncle who is feeling sick."
+    print(f"Intent: {classify_intent(sample_text)}")

vocca_ai/preprocess.py

@@ -0,0 +1,4 @@
+def priority_score(text):
+    urgent_keywords = ["emergency", "urgent", "severe", "immediate", "pain", "critical"]
+    score = sum(1 for word in text.lower().split() if word in urgent_keywords)
+    return "High" if score > 1 else "Medium" if score == 1 else "Low"

vocca_ai/sentiment.py

@@ -0,0 +1,21 @@
+from transformers import pipeline
+
+# Load a better sentiment model
+sentiment_model = pipeline("sentiment-analysis", model="cardiffnlp/twitter-xlm-roberta-base-sentiment")
+
+def analyze_sentiment(text):
+    """
+    Uses a specialized sentiment model better suited for medical text.
+    """
+    result = sentiment_model(text)[0]["label"]
+
+    if result.lower() == "positive":
+        return "Positive"
+    elif result.lower() == "negative":
+        return "Concerned"  # Replace "Negative" with a more medically appropriate term
+    return "Neutral"
+
+# Example Usage
+if __name__ == "__main__":
+    sample_text = "I've been feeling really weak and dizzy for the past few days."
+    print(f"Sentiment: {analyze_sentiment(sample_text)}")