Update app.py

app.py

@@ -2,6 +2,7 @@ import gradio as gr
 import google.generativeai as genai
 import os
 import textwrap
+import re
 from dotenv import load_dotenv
 
 # RAG Specific Imports
@@ -21,12 +22,12 @@ except (ValueError, AttributeError) as e:
     print(f"⚠️ Configuration Error: {e}")
     exit()
 
-# --- RAG Setup (Runs once on startup) ---
+# --- RAG Setup Function ---
+# This function will now be called directly, only once.
 def setup_rag_pipeline():
     global retriever
     print("Setting up RAG pipeline...")
     try:
-        # Make sure your final optimized source file is named this
         input_file = 'filtered_gardening_kb.txt'
         with open(input_file, 'r', encoding='utf-8') as f:
             source_text = f.read()
@@ -36,10 +37,11 @@ def setup_rag_pipeline():
         documents = text_splitter.create_documents([source_text])
         
         print(f"Creating embeddings for {len(documents)} chunks...")
+        # Note: Embedding has a different, much higher free quota than generation.
         embeddings = GoogleGenerativeAIEmbeddings(model="models/embedding-001")
         
         vector_store = FAISS.from_documents(documents, embeddings)
-        retriever = vector_store.as_retriever(search_kwargs={"k": 7}) # Increased to 7 chunks for more context
+        retriever = vector_store.as_retriever(search_kwargs={"k": 7})
         
         print("✅ RAG pipeline setup complete!")
 
@@ -50,35 +52,27 @@ def setup_rag_pipeline():
         print(f"An error occurred during RAG setup: {e}")
         retriever = None
 
-# --- Gemini Model and NEW, MORE LENIENT PROMPT ---
+# --- Gemini Model and Prompt ---
 MODEL_NAME = 'gemini-1.5-flash-latest'
 
-# UPDATED: This prompt encourages a more natural, expert-like tone.
 SYSTEM_PROMPT = textwrap.dedent("""
     You are a knowledgeable and friendly gardening expert from a community chat group. Your entire knowledge base comes from the CONTEXT CHUNKS provided below, which are excerpts from your group's conversations.
-
     Your goal is to answer the user's questions in a helpful, conversational, and natural way, as if you are sharing your collective experience.
-
     **Your Guidelines:**
-    1.  **Base all answers on the source:** Your responses must be directly inspired by and grounded in the information within the CONTEXT CHUNKS. Prioritize the advice, techniques, and problems discussed in the text.
-    2.  **Synthesize and Connect:** Feel free to connect related ideas from different parts of the context to provide a more complete and helpful answer. Do not just repeat snippets.
-    3.  **Adopt a helpful persona:** Sound like an experienced member of a gardening group sharing advice, not a robot querying a database. Use a natural, encouraging tone.
-    4.  **Handle missing information gracefully:** If the user's question is completely outside the scope of the provided text, you can say something like, "That's an interesting question, but our group chats don't seem to cover that specific topic." If it's related but not explicit, you can say, "While the chats don't mention that directly, the general advice on [related topic] suggests that..."
-    5.  **Do NOT add citations:** Do not include markers like `[1]` or citation separators. Just provide a clean, direct answer.
+    1.  Base all answers on the source: Your responses must be directly inspired by and grounded in the information within the CONTEXT CHUNKS.
+    2.  Synthesize and Connect: Feel free to connect related ideas from different parts of the context to provide a more complete answer.
+    3.  Adopt a helpful persona: Sound like an experienced member of a gardening group sharing advice.
+    4.  Handle missing information gracefully: If the user's question is completely outside the scope of the provided text, state that the chats don't cover that topic. If it's related but not explicit, you can say, "While the chats don't mention that directly, the general advice on [related topic] suggests that..."
+    5.  Do NOT add citations: Just provide a clean, direct answer.
 """).strip()
 
-
-# --- UPDATED: Simplified Response Generation Function ---
 def generate_gemini_response(message, chat_history):
-    """Generates the chatbot's response using the RAG pipeline."""
     if not retriever:
-        return "The RAG pipeline is not initialized. Please check the application logs for errors."
+        return "The RAG pipeline is not initialized. Please check application logs."
 
-    # 1. Retrieve relevant documents
     relevant_docs = retriever.invoke(message)
     context_chunks = "\n---\n".join([doc.page_content for doc in relevant_docs])
 
-    # 2. Construct the prompt with the retrieved context
     full_prompt = f"""{SYSTEM_PROMPT}
 
 <CONTEXT CHUNKS>
@@ -90,15 +84,15 @@ QUERY: {message}
     try:
         model = genai.GenerativeModel(MODEL_NAME)
         response = model.generate_content(full_prompt)
-        
-        # The entire citation parsing block is removed. We just return the clean text.
         return response.text.strip()
-
     except Exception as e:
+        # Provide a more user-friendly message for quota errors
+        if "quota" in str(e).lower():
+            return "The daily free API quota has been reached. Please try again tomorrow or upgrade your Google AI plan by enabling billing."
         print(f"Error generating response: {e}")
-        return "An error occurred while communicating with the AI model. Please check the API key and try again."
+        return "An error occurred while communicating with the AI model."
 
-# --- Gradio UI & Styling (No changes needed) ---
+# --- Gradio UI & Styling ---
 CSS = """
 /* (Your existing CSS code remains unchanged here) */
 @import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap');
@@ -120,6 +114,11 @@ body, html { font-family: 'Inter', sans-serif; background-color: #121212; color:
 #send-button:hover { background-color: #0056b3; transform: translateY(-1px); }
 """
 
+# --- Main Application Logic ---
+# First, set up the RAG pipeline globally.
+retriever = None
+setup_rag_pipeline()
+
 with gr.Blocks(css=CSS, theme=gr.themes.Base()) as demo:
     with gr.Column(elem_id="main-container"):
         with gr.Row(elem_id="header"):
@@ -157,8 +156,9 @@ with gr.Blocks(css=CSS, theme=gr.themes.Base()) as demo:
 
     btn_submit.click(process_and_clear, [txt_msg, chatbot], [chatbot, txt_msg])
     txt_msg.submit(process_and_clear, [txt_msg, chatbot], [chatbot, txt_msg])
-    
-    demo.load(setup_rag_pipeline, None, None)
+
+# The demo.load() is removed, as setup is now handled before the Blocks are defined.
 
 if __name__ == "__main__":
-    demo.queue().launch(debug=True)
+    # For deploying on Hugging Face, it's better to set debug=False
+    demo.queue().launch(debug=False)