Update app.py

app.py

@@ -1,44 +1,27 @@
 import gradio as gr
 import pandas as pd
 import os
+import random
 from PIL import Image
+import torch
 from torchvision import transforms
-from langchain_community.embeddings import HuggingFaceEmbeddings
-from langchain_groq import ChatGroq
 
 # --- Load CSV Data ---
 def load_treatments_data():
     try:
         context = pd.read_csv('./crop_diseases_treatments.csv')
         print("CSV file loaded successfully.")
-        context_data = []
-        for i in range(len(context)):
-            contextn = ""
-            for j in range(len(context.columns)):
-                contextn += context.columns[j] + ":" + str(context.iloc[i][j]) + " "
-            context_data.append(contextn)
-        return context, context_data
+        return context
     except FileNotFoundError:
         print("Error: crop_diseases_treatments.csv not found.")
         # Create a minimal dataframe for demonstration
         return pd.DataFrame({
-            'Crop': ['Tomato', 'Apple'],
-            'Disease': ['Early Blight', 'Apple Scab'],
-            'Symptoms': ['Yellow spots on leaves', 'Dark scab-like lesions'],
-            'Treatment': ['Remove affected leaves', 'Prune affected branches'],
-            'Medicine/Chemical Control': ['Copper fungicide', 'Sulfur spray']
-        }), ["Error: crop_diseases_treatments.csv not found."]
-
-# --- Load Groq LLM ---
-def initialize_llm():
-    groq_key = os.environ.get('GROQ_API_KEY')
-    if not groq_key:
-        print("Warning: GROQ_API_KEY not found in environment variables. The chatbot may not work.")
-    return ChatGroq(model="llama-3.3-70b-versatile", api_key=groq_key)
-
-# --- Load Embedding Model ---
-def initialize_embedding_model():
-    return HuggingFaceEmbeddings(model_name="mixedbread-ai/mxbai-embed-large-v1")
+            'Crop': ['Tomato', 'Apple', 'Corn', 'Rice', 'Potato'],
+            'Disease': ['Early Blight', 'Apple Scab', 'Common Rust', 'Bacterial Leaf Blight', 'Late Blight'],
+            'Symptoms': ['Yellow spots on leaves', 'Dark scab-like lesions', 'Rust-colored pustules', 'Yellow-orange lesions', 'Dark water-soaked spots'],
+            'Treatment': ['Remove affected leaves', 'Prune affected branches', 'Remove infected plants', 'Drain fields', 'Apply fungicide early'],
+            'Medicine/Chemical Control': ['Copper fungicide', 'Sulfur spray', 'Propiconazole', 'Streptomycin sulfate', 'Chlorothalonil']
+        })
 
 # --- Disease Classification ---
 def prepare_disease_classifier():
@@ -71,10 +54,7 @@ def classify_disease(image):
     
     # Process the image
     if image is not None:
-        img_tensor = transform(image).unsqueeze(0)
-        
         # For demonstration, we'll randomly select a class
-        import random
         class_idx = random.randint(0, len(class_names) - 1)
         class_name = class_names[class_idx]
         
@@ -132,41 +112,102 @@ def find_treatment(crop, disease, df):
             "medicine": "Consult a plant pathologist"
         }
 
-# Function to answer questions using Groq
-def answer_question_with_groq(question, df, llm):
-    # Create context from the database
-    context = ""
-    for _, row in df.iterrows():
-        context += f"Crop: {row['Crop']}, Disease: {row['Disease']}, "
-        context += f"Symptoms: {row['Symptoms']}, Treatment: {row['Treatment']}, "
-        context += f"Medicine: {row['Medicine/Chemical Control']}\n"
+# Simple function to answer questions based on the database
+def answer_question(question, df):
+    question = question.lower()
+    
+    # Look for crop and disease mentions in the question
+    crop_match = None
+    disease_match = None
+    
+    for crop in df['Crop'].unique():
+        if crop.lower() in question:
+            crop_match = crop
+            break
     
-    # Prepare the prompt for Groq
-    prompt = f"""
-    You are a plant disease expert. Answer the following question about plant diseases and treatments.
-    Use only the information provided in the context. If you don't know the answer, say so.
+    for disease in df['Disease'].unique():
+        if disease.lower() in question:
+            disease_match = disease
+            break
     
-    Context:
-    {context[:4000]}  # Limit context size
+    # If we found both crop and disease
+    if crop_match and disease_match:
+        matches = df[(df['Crop'] == crop_match) & (df['Disease'] == disease_match)]
+        if not matches.empty:
+            match = matches.iloc[0]
+            return f"""
+            Crop: {match['Crop']}
+            Disease: {match['Disease']}
+            Symptoms: {match['Symptoms']}
+            Treatment: {match['Treatment']}
+            Medicine: {match['Medicine/Chemical Control']}
+            """
+    
+    # If we found only crop
+    elif crop_match:
+        matches = df[df['Crop'] == crop_match]
+        if not matches.empty:
+            response = f"Information about diseases affecting {crop_match}:\n\n"
+            for _, row in matches.iterrows():
+                response += f"Disease: {row['Disease']}\n"
+                response += f"Symptoms: {row['Symptoms']}\n"
+                response += f"Treatment: {row['Treatment']}\n"
+                response += f"Medicine: {row['Medicine/Chemical Control']}\n\n"
+            return response
     
-    Question: {question}
+    # If we found only disease
+    elif disease_match:
+        matches = df[df['Disease'] == disease_match]
+        if not matches.empty:
+            response = f"Information about {disease_match} affecting various crops:\n\n"
+            for _, row in matches.iterrows():
+                response += f"Crop: {row['Crop']}\n"
+                response += f"Symptoms: {row['Symptoms']}\n"
+                response += f"Treatment: {row['Treatment']}\n"
+                response += f"Medicine: {row['Medicine/Chemical Control']}\n\n"
+            return response
     
-    Answer with the following format:
-    Crop: [crop name if identified]
-    Disease: [disease name if identified]
-    Treatment: [treatment recommendations]
-    Medicine: [chemical control options]
-    """
+    # General search
+    keywords = question.split()
+    relevant_rows = []
     
-    try:
-        # Get response from Groq
-        response = llm.invoke(prompt)
-        return response.content
-    except Exception as e:
-        return f"Error getting response from Groq: {str(e)}\n\nPlease check your API key and try again."
+    for _, row in df.iterrows():
+        score = 0
+        for keyword in keywords:
+            if len(keyword) < 3:  # Skip short words
+                continue
+            if keyword in str(row['Crop']).lower():
+                score += 3
+            if keyword in str(row['Disease']).lower():
+                score += 5
+            if keyword in str(row['Symptoms']).lower():
+                score += 2
+            if keyword in str(row['Treatment']).lower():
+                score += 1
+            if keyword in str(row['Medicine/Chemical Control']).lower():
+                score += 1
+        
+        if score > 0:
+            relevant_rows.append((score, row))
+    
+    if relevant_rows:
+        # Sort by relevance score
+        relevant_rows.sort(key=lambda x: x[0], reverse=True)
+        top_matches = relevant_rows[:3]  # Get top 3 matches
+        
+        response = "Here's what I found based on your question:\n\n"
+        for _, row in top_matches:
+            response += f"Crop: {row['Crop']}\n"
+            response += f"Disease: {row['Disease']}\n"
+            response += f"Symptoms: {row['Symptoms']}\n"
+            response += f"Treatment: {row['Treatment']}\n"
+            response += f"Medicine: {row['Medicine/Chemical Control']}\n\n"
+        return response
+    
+    return "I couldn't find specific information related to your question in the database. Please try asking about specific crops or diseases."
 
 # Function to process image uploads
-def process_image(image, df, llm):
+def process_image(image, df):
     if image is None:
         return "Please upload an image to analyze.", None, None, None, None
     
@@ -176,37 +217,20 @@ def process_image(image, df, llm):
     # Get treatment information
     treatment_info = find_treatment(crop, disease, df)
     
-    # Use Groq to enhance the analysis
-    image_description = f"Image shows a {crop} plant with possible {disease} disease."
-    prompt = f"""
-    You are a plant disease expert. Analyze this plant image description and provide detailed information.
-    
-    Image description: {image_description}
-    
-    Based on the database information, this appears to be {disease} affecting {crop}.
-    
-    Provide a detailed analysis with the following format:
-    Crop: {crop}
-    Disease: {disease}
-    Symptoms: [describe typical symptoms]
-    Treatment: [provide treatment recommendations]
-    Medicine: [suggest chemical control options]
-    """
+    # Create analysis text
+    if disease.lower() == "healthy":
+        analysis = f"Good news! This {crop} plant appears to be healthy."
+    else:
+        analysis = f"Analysis: This {crop} plant appears to be affected by {disease}.\n\n"
+        analysis += f"Typical symptoms of {disease} include {treatment_info['symptoms']}.\n\n"
+        analysis += f"Recommended treatment: {treatment_info['treatment']}"
     
-    try:
-        # Get enhanced analysis from Groq
-        groq_analysis = llm.invoke(prompt)
-        return groq_analysis.content, treatment_info["crop"], treatment_info["disease"], treatment_info["treatment"], treatment_info["medicine"]
-    except Exception as e:
-        # Fallback to basic information if Groq fails
-        return f"Analysis: {crop} affected by {disease}", treatment_info["crop"], treatment_info["disease"], treatment_info["treatment"], treatment_info["medicine"]
+    return analysis, treatment_info["crop"], treatment_info["disease"], treatment_info["treatment"], treatment_info["medicine"]
 
 # Main function to set up the Gradio interface
 def main():
-    # Load data and initialize models
-    df, context_data = load_treatments_data()
-    llm = initialize_llm()
-    embed_model = initialize_embedding_model()
+    # Load data
+    df = load_treatments_data()
     
     # Create a simplified Gradio interface
     with gr.Blocks(title="Plant Disease Assistant") as app:
@@ -244,13 +268,13 @@ def main():
         
         # Set up event handlers
         image_submit.click(
-            fn=lambda img: process_image(img, df, llm),
+            fn=lambda img: process_image(img, df),
             inputs=[image_input],
             outputs=[analysis_output, crop_output, disease_output, treatment_output, medicine_output]
         )
         
         question_submit.click(
-            fn=lambda q: (answer_question_with_groq(q, df, llm), "", "", "", ""),
+            fn=lambda q: (answer_question(q, df), "", "", "", ""),
             inputs=[question_input],
             outputs=[analysis_output, crop_output, disease_output, treatment_output, medicine_output]
         )