Update app.py

app.py

@@ -12,10 +12,11 @@ import gradio as gr
 from docx import Document
 import PyPDF2
 import csv
+import google.generativeai as genai
 
 warnings.filterwarnings("ignore")
+global context
 
-# Function to extract text from PDF
 def extract_text_from_pdf(pdf_path):
     with open(pdf_path, "rb") as f:
         pdf_reader = PyPDF2.PdfReader(f)
@@ -25,7 +26,6 @@ def extract_text_from_pdf(pdf_path):
             text += page.extract_text()
         return text
 
-# Function to extract text from DOCX
 def extract_text_from_docx(docx_path):
     doc = Document(docx_path)
     full_text = []
@@ -33,7 +33,6 @@ def extract_text_from_docx(docx_path):
         full_text.append(para.text)
     return '\n\n'.join(full_text)
 
-# Function to extract text from TXT
 def extract_text_from_txt(txt_path):
     try:
         with open(txt_path, "r", encoding='utf-8') as f:
@@ -42,7 +41,17 @@ def extract_text_from_txt(txt_path):
         with open(txt_path, "r", encoding='latin-1') as f:
             return f.read()
     
+def extract_text_from_json(json_path):
+    with open(json_path, "r", encoding='utf-8') as f:
+        try:
+            data = json.load(f)
+            if not data:  
+                return ""
+            return json.dumps(data, indent=4)
+        except json.JSONDecodeError:
+            return ""
 
+    
 def read_and_structure_csv(csv_path):
     structured_data = []
     with open(csv_path, mode='r', encoding='utf-8-sig') as file:
@@ -54,33 +63,30 @@ def read_and_structure_csv(csv_path):
                     plan_details += f"  - **{key.replace('_', ' ').title()}**: {value}\n"
             structured_data.append(plan_details)
     return "\n\n".join(structured_data)
-# List of paths to your input files (PDF, DOCX, TXT)
+
 file_paths = ["./csvrecommend - Sheet1.csv","./dummymedicare.txt"]
 
-# Extract text from files
-texts = []
+texts1 = []
 for path in file_paths:
     if path.endswith(".pdf"):
-        texts.append(extract_text_from_pdf(path))
+        texts1.append(extract_text_from_pdf(path))
     elif path.endswith(".docx"):
-        texts.append(extract_text_from_docx(path))
+        texts1.append(extract_text_from_docx(path))
     elif path.endswith(".txt"):
         txt_content = extract_text_from_txt(path)
-        texts.append(txt_content)
-        print("TXT file content:", txt_content)
-        print(f"Added content from {path}: {txt_content[:500]}...")
+        texts1.append(txt_content)
+        
     elif path.endswith(".csv"):
-        texts.append(read_and_structure_csv(path))
+        texts1.append(read_and_structure_csv(path))
+    elif path.endswith(".json"):
+        texts1.append(extract_text_from_json(path))
 
-context = "\n\n".join(texts)
 
-print(context,"context\n\n\n")
+context = "\n\n".join(texts1)
 
-# Initialize text splitter and embeddings
-text_splitter = RecursiveCharacterTextSplitter(chunk_size=8000, chunk_overlap=1200)
+text_splitter = RecursiveCharacterTextSplitter(chunk_size=8000, chunk_overlap=1300)
 texts = text_splitter.split_text(context)
 
-print("\n\n\n",texts,"texts")
 api_key = "AIzaSyCqEKwd23ztVuk-dkCXypjeHWlcs41aCSM"
 if not api_key:
     raise ValueError("API key not found. Please set your GEMINI_API_KEY in the environment.")
@@ -95,7 +101,7 @@ embeddings = GoogleGenerativeAIEmbeddings(model="models/embedding-001", google_a
 vector_index = Chroma.from_texts(texts, embeddings).as_retriever(search_kwargs={"k": 5})
 
 # Create QA chain
-template = """You are a highly knowledgeable and detail-oriented medical assistant specializing in insurance plans. Your task is to recommend only those insurance plans that strictly align with all the needs and preferences provided by the user.
+template = """You are a highly knowledgeable and detail-oriented medical assistant specializing in insurance plans. Your task is to answer questions to user and recommend only those insurance plans that strictly align with all the needs and preferences provided by the user.
 Ensure that each recommended plan meets every single requirement specified by the user. Do not recommend plans that only partially meet the requirements.
 Context:
 {context}
@@ -110,18 +116,17 @@ qa_chain = RetrievalQA.from_chain_type(
     chain_type_kwargs={"prompt": QA_CHAIN_PROMPT}
 )
 
-# Load history from file if it exists
-history_file = "history.json"
+history_file = "history2.json"
 
 def load_history():
     if os.path.exists(history_file):
         with open(history_file, "r") as f:
             try:
                 data = json.load(f)
-                if isinstance(data, list):  # Ensure that the data is a list
+                if isinstance(data, list):  
                     return data
             except json.JSONDecodeError:
-                pass  # If there's an error decoding, return an empty list
+                pass  
     return []
 
 def save_history(history):
@@ -130,32 +135,55 @@ def save_history(history):
 
 history = load_history()
 
-# Function to handle questions and maintain history
+def summarize_history(history):
+    os.environ['GOOGLE_API_KEY'] = "AIzaSyCqEKwd23ztVuk-dkCXypjeHWlcs41aCSM"
+    genai.configure(api_key = os.environ['GOOGLE_API_KEY'])
+    # Summarize the user's preferences from the chat history
+    user_history = "\n".join([entry["USER"] for entry in history])
+    prompt = f"Summarize the important points  on preferences from the following user history Discard unwanted details only give users preferences :\n\n{user_history}"
+    model1=genai.GenerativeModel('gemini-pro')
+
+    summary_response = model1.generate_content(prompt)
+    print(summary_response.text,"summary_response")
+    summary = summary_response.text
+    return summary
+
 def ask_question(question):
     global history
     if question.strip().lower() == "exit":
-        history = []  # Clear history
+        history = []  
         save_history(history)
         return "Hey there! I'm your Medicare assistant. You can ask me questions related to different type of insurances and I'll help you. Let's get started!"
+    
+    with open("./chat_history.txt", "a") as f:
+        f.write(f"USER: {question}\n")
+
     result = qa_chain({"query": question})
     answer = result["result"]
-    history.append({"question": question, "answer": answer})
+    history.append({"USER": question, "answer": answer})
     save_history(history)
-    # Format history for display
+
+    # Summarize the chat history
+    summary = summarize_history(history)
+    
+    # Combine the context and summary for the text splitter
+    combined_text = context + "\n\n" + "MY PREFERENCES "+summary
+    print(combined_text,"combined_text\n\n")
+    texts = text_splitter.split_text(combined_text)
+    vector_index = Chroma.from_texts(texts, embeddings).as_retriever(search_kwargs={"k": 5})
+
     history_md = ""
     for entry in history:
-        history_md += f"**USER:** {entry['question']}\n\n**BOT:** {entry['answer']}\n\n---\n\n"
+        history_md += f"**USER:** {entry['USER']}\n\n**BOT:** {entry['answer']}\n\n---\n\n"
     return history_md
 
-# Format history for initial display
 initial_history_md = ""
 if not history:
     initial_history_md = "Hey there! I'm your Medicare assistant. You can ask me questions related to different type of insurances and I'll help you. Let's get started!"
 else:
     for entry in history:
-        initial_history_md += f"**User:** {entry['question']}\n\n**Bot:** {entry['answer']}\n\n---\n\n"
+        initial_history_md += f"**USER:** {entry['USER']}\n\n**BOT:** {entry['answer']}\n\n---\n\n"
 
-# Create Gradio interface using Blocks
 with gr.Blocks() as demo:
     gr.HTML(
         """
@@ -177,22 +205,15 @@ with gr.Blocks() as demo:
         """
     )
     
-    # Markdown block to display history
     history_output = gr.Markdown(value=initial_history_md, elem_classes="scrollable-history")
     
-    # Row for question input and submit button
     with gr.Row(elem_classes="fixed-bottom"):
-        # Column for question input and submit button
         with gr.Column():
-            # Text area for question input
             question_input = gr.Textbox(lines=2, placeholder="Type your question here...", show_label=False)
-            # Submit button
             submit_button = gr.Button("Submit")
-            # Function to handle submit action
             submit_button.click(ask_question, inputs=question_input, outputs=history_output)
-            submit_button.click(lambda: "", None, question_input)  # Clears the input box
+            submit_button.click(lambda: "", None, question_input)  
     
-    # Display history above the question input and submit button pair
     history_output
 
 demo.launch()