Update app.py

app.py

@@ -1,12 +1,129 @@
-print("--- app.py: TOP OF FILE ---")
+# app.py (DEBUGGING VERSION)
+
+print("--- Python script starting ---")
+
 import streamlit as st
-print("--- app.py: Streamlit imported ---")
 import os
-print("--- app.py: os imported ---")
-
-st.set_page_config(page_title="Clean Slate Test", page_icon="🧼")
-st.title("🧼 Clean Slate Test App")
-st.write("If you see this, the basic Streamlit environment on a FRESH SPACE is working.")
-print("--- app.py: Streamlit UI rendered ---")
-print(f"CWD: {os.getcwd()}")
-print("--- app.py: END OF FILE ---")
+from dotenv import load_dotenv
+from pinecone import Pinecone
+
+# --- Standard Imports ---
+from langchain_pinecone import PineconeVectorStore
+from langchain_community.embeddings import SentenceTransformerEmbeddings
+from langchain_groq import ChatGroq
+from langchain_core.prompts import PromptTemplate
+from langchain_core.runnables import RunnablePassthrough
+from langchain_core.output_parsers import PydanticOutputParser
+from pydantic import BaseModel, Field
+from langchain.retrievers import ContextualCompressionRetriever
+from langchain.retrievers.document_compressors import CohereRerank
+
+print("--- All imports successful ---")
+
+# We wrap the ENTIRE app in a try/except block to catch any startup error
+try:
+    # --- Load Environment Variables ---
+    print("Step 1: Loading environment variables...")
+    load_dotenv()
+    PINECONE_API_KEY = os.getenv('PINECONE_API_KEY')
+    GROQ_API_KEY = os.getenv('GROQ_API_KEY')
+    COHERE_API_KEY = os.getenv('COHERE_API_KEY')
+    INDEX_NAME = "rag-chatbot"
+    print("Step 1: SUCCESS")
+
+    # --- Page Configuration ---
+    st.set_page_config(page_title="Production RAG System", page_icon="🚀", layout="wide")
+    st.title("🚀 Production-Grade RAG System")
+
+    # --- Pydantic Model ---
+    class StructuredAnswer(BaseModel):
+        summary: str = Field(description="A concise summary.")
+        key_points: list[str] = Field(description="A list of key bullet points.")
+        confidence_score: float = Field(description="A 0.0 to 1.0 confidence score.")
+
+    # --- Caching and Initialization ---
+    @st.cache_resource
+    def initialize_services():
+        print("Step 2: Entering initialize_services function...")
+        if not all([PINECONE_API_KEY, GROQ_API_KEY, COHERE_API_KEY]):
+            raise ValueError("An API key is missing!")
+
+        print("Step 2a: Initializing embedding model...")
+        embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
+        print("Step 2a: SUCCESS")
+
+        print("Step 2b: Initializing Pinecone client...")
+        pinecone = Pinecone(api_key=PINECONE_API_KEY)
+        host = "https://rag-chatbot-sg8t88c.svc.aped-4627-b74a.pinecone.io"
+        index = pinecone.Index(host=host)
+        print("Step 2b: SUCCESS")
+
+        print("Step 2c: Creating PineconeVectorStore object...")
+        vectorstore = PineconeVectorStore(index=index, embedding=embeddings)
+        print("Step 2c: SUCCESS")
+
+        print("Step 2d: Initializing Cohere Re-ranker...")
+        base_retriever = vectorstore.as_retriever(search_kwargs={'k': 20})
+        compressor = CohereRerank(cohere_api_key=COHERE_API_KEY, top_n=5)
+        reranking_retriever = ContextualCompressionRetriever(base_compressor=compressor, base_retriever=base_retriever)
+        print("Step 2d: SUCCESS")
+
+        print("Step 2e: Initializing Groq LLM...")
+        llm = ChatGroq(temperature=0, model_name="llama3-70b-8192", api_key=GROQ_API_KEY)
+        print("Step 2e: SUCCESS")
+
+        print("Step 2: All services initialized successfully.")
+        return reranking_retriever, llm
+
+    print("Step 3: Calling initialize_services...")
+    retriever, llm = initialize_services()
+    print("Step 3: SUCCESS, services are loaded.")
+
+    # --- RAG Chain Definition ---
+    print("Step 4: Defining RAG chain...")
+    pydantic_parser = PydanticOutputParser(pydantic_object=StructuredAnswer)
+    format_instructions = pydantic_parser.get_format_instructions()
+    template = """
+    You are a world-class analysis engine. Your task is to provide a structured, factual answer based *only* on the following context.
+    Synthesize the information from all context snippets. Do not use any outside knowledge.
+
+    Context:
+    {context}
+
+    Question:
+    {question}
+
+    Follow these formatting instructions precisely:
+    {format_instructions}
+    """
+    prompt = PromptTemplate(
+        template=template,
+        input_variables=["context", "question"],
+        partial_variables={"format_instructions": format_instructions}
+    )
+    rag_chain = (
+        {"context": retriever, "question": RunnablePassthrough()}
+        | prompt
+        | llm
+        | pydantic_parser
+    )
+    print("Step 4: SUCCESS")
+
+    # --- UI Rendering ---
+    print("Step 5: Starting to render Streamlit UI...")
+    st.success("System is ready. Ask your question below.")
+    query = st.text_input("Enter your question:", key="query_input")
+
+    if query:
+        with st.spinner("Processing..."):
+            structured_answer = rag_chain.invoke(query)
+            st.write("### Answer")
+            # ... rest of UI ...
+    print("Step 5: SUCCESS, UI is rendered.")
+
+except Exception as e:
+    # If ANY error happens during startup, it will be printed here
+    print(f"!!!!!!!!!! A FATAL ERROR OCCURRED !!!!!!!!!!")
+    import traceback
+    print(traceback.format_exc())
+    st.error(f"A fatal error occurred during startup. Please check the container logs. Error: {e}")