Update app.py

app.py

@@ -0,0 +1,216 @@
+# app.py
+from langchain_groq import ChatGroq
+from langchain.vectorstores import Chroma
+from langchain.document_loaders import PyPDFLoader, DirectoryLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.embeddings import HuggingFaceBgeEmbeddings
+from transformers import pipeline
+import gradio as gr
+import os
+
+# Initialize Groq with environment variable
+llm = ChatGroq(
+    temperature=0.7,
+    groq_api_key=os.environ.get("Groq_API_KEY"),  # Set in HF Secrets
+    model_name="llama-3.3-70b-versatile"
+)
+
+# Configure paths for Hugging Face Space
+VECTOR_DB_PATH = "./chroma_db"
+PDF_DIR = "./Pregnancy"
+MODEL_PATH = "./indian_food_model"
+
+# Initialize or load vector database
+if not os.path.exists(VECTOR_DB_PATH):
+    # Create new vector database
+    loader = DirectoryLoader(PDF_DIR, glob="*.pdf", loader_cls=PyPDFLoader)
+    documents = loader.load()
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
+    texts = text_splitter.split_documents(documents)
+    embeddings = HuggingFaceBgeEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
+    vector_db = Chroma.from_documents(texts, embeddings, persist_directory=VECTOR_DB_PATH)
+else:
+    # Load existing database
+    embeddings = HuggingFaceBgeEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
+    vector_db = Chroma(persist_directory=VECTOR_DB_PATH, embedding_function=embeddings)
+
+retriever = vector_db.as_retriever()
+
+# Load food classification model
+food_classifier = pipeline(
+    "image-classification", 
+    model=MODEL_PATH,
+    device_map="auto"
+)
+
+def classify_food(image):
+    """Classify food images with confidence thresholding"""
+    if image is None:
+        return None, 0.0
+    results = food_classifier(image)
+    if not results:
+        return None, 0.0
+    top_result = results[0]
+    label = top_result["label"]
+    score = top_result["score"]
+    if score < 0.3 or "non-food" in label.lower():
+        return None, score
+    return label, score
+
+def format_history(chat_history, max_exchanges=5):
+    """Format conversation history for context"""
+    recent_history = chat_history[-max_exchanges:]
+    return "\n".join(
+        f"User: {user}\nAssistant: {assistant}" 
+        for user, assistant in recent_history
+    )
+
+def calculate_metrics(status, pre_weight, current_weight, height, 
+                     gest_age=None, time_since_delivery=None, breastfeeding=None):
+    """Calculate pregnancy/postpartum metrics"""
+    if None in [pre_weight, current_weight, height]:
+        return "Missing required fields: weight and height"
+    
+    height_m = height / 100
+    pre_bmi = pre_weight / (height_m ** 2)
+    
+    if status == "Pregnant":
+        if not gest_age or not (0 <= gest_age <= 40):
+            return "Invalid gestational age (0-40 weeks)"
+        
+        # BMI-based recommendations
+        bmi_ranges = [
+            (18.5, 12.5, 18),
+            (25, 11.5, 16),
+            (30, 7, 11.5),
+            (float('inf'), 5, 9)
+        ]
+        for max_bmi, min_gain, max_gain in bmi_ranges:
+            if pre_bmi < max_bmi:
+                break
+                
+        current_gain = current_weight - pre_weight
+        expected_min = (min_gain / 40) * gest_age
+        expected_max = (max_gain / 40) * gest_age
+        
+        if current_gain < expected_min:
+            advice = "Consider nutritional counseling"
+        elif current_gain > expected_max:
+            advice = "Consult your healthcare provider"
+        else:
+            advice = "Good progress! Maintain balanced diet"
+            
+        return (f"Pre-BMI: {pre_bmi:.1f}\nWeek {gest_age} recommendation: "
+                f"{expected_min:.1f}-{expected_max:.1f} kg\n"
+                f"Your gain: {current_gain:.1f} kg\n{advice}")
+    
+    elif status == "Postpartum":
+        if None in [time_since_delivery, breastfeeding]:
+            return "Missing postpartum details"
+        
+        current_bmi = current_weight / (height_m ** 2)
+        if breastfeeding == "Yes":
+            advice = ("Aim for 0.5-1 kg/month loss while breastfeeding\n"
+                     "Focus on nutrient-dense foods")
+        else:
+            advice = "Gradual weight loss through diet and exercise"
+            
+        return (f"Current BMI: {current_bmi:.1f}\n"
+               f"{time_since_delivery} weeks postpartum\n{advice}")
+    
+    return "Select pregnancy status"
+
+def chat_function(user_input, image, chat_history):
+    """Handle chat interactions"""
+    history_str = format_history(chat_history)
+    
+    # Crisis detection
+    crisis_terms = {
+        "suicide", "self-harm", "kill myself", "hopeless",
+        "panic attack", "worthless", "end it all"
+    }
+    if any(term in user_input.lower() for term in crisis_terms):
+        return chat_history + [(user_input, crisis_response)]
+    
+    # Process image or text input
+    if image:
+        food_label, confidence = classify_food(image)
+        if food_label:
+            context = f"User uploaded {food_label} image. {user_input}"
+        else:
+            return chat_history + [(user_input, "Couldn't identify food in image")]
+    else:
+        context = user_input
+    
+    # Retrieve relevant documents
+    docs = retriever.get_relevant_documents(context)
+    context_str = "\n".join(d.page_content for d in docs[:3])
+    
+    # Generate response
+    prompt = f"""Context from documents:
+{context_str}
+
+Recent conversation:
+{history_str}
+
+User: {context}
+Assistant:"""
+    
+    response = llm.invoke(prompt).content
+    return chat_history + [(user_input, response)]
+
+# Crisis response template
+crisis_response = """🚨 Immediate Help Resources:
+- India: Vandrevala Foundation - 1860 266 2345
+- US: 988 Suicide & Crisis Lifeline
+- UK: Samaritans - 116 123
+- Worldwide: https://findahelpline.com"""
+
+# Gradio Interface
+with gr.Blocks(css=custom_css, theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 🤰 Maternal Wellness Companion")
+    
+    with gr.Row():
+        with gr.Column(scale=2):
+            chatbot = gr.Chatbot(height=500)
+            input_txt = gr.Textbox(placeholder="Ask about pregnancy health...")
+            input_img = gr.Image(type="pil", label="Upload Food Image")
+            btn_send = gr.Button("Send", variant="primary")
+            
+        with gr.Column(scale=1):
+            gr.Markdown("## Health Metrics")
+            status = gr.Radio(["Pregnant", "Postpartum"], label="Status")
+            pre_weight = gr.Number(label="Pre-pregnancy Weight (kg)")
+            current_weight = gr.Number(label="Current Weight (kg)")
+            height = gr.Number(label="Height (cm)")
+            gest_age = gr.Number(visible=False)
+            postpartum_time = gr.Number(visible=False)
+            breastfeeding = gr.Radio(["Yes", "No"], visible=False)
+            btn_calculate = gr.Button("Calculate", variant="secondary")
+            
+    # Event handling
+    status.change(
+        lambda s: (
+            gr.update(visible=s=="Pregnant"),
+            gr.update(visible=s=="Postpartum"),
+            gr.update(visible=s=="Postpartum")
+        ),
+        inputs=status,
+        outputs=[gest_age, postpartum_time, breastfeeding]
+    )
+    
+    btn_send.click(
+        chat_function,
+        [input_txt, input_img, chatbot],
+        [chatbot],
+        queue=False
+    )
+    
+    btn_calculate.click(
+        calculate_metrics,
+        [status, pre_weight, current_weight, height, 
+         gest_age, postpartum_time, breastfeeding],
+        chatbot
+    )
+
+demo.launch(debug=False)