app.py

import os
import streamlit as st
import pickle
import time
import requests
from bs4 import BeautifulSoup

# ---- LangChain Community Packages ----
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain.chains import RetrievalQAWithSourcesChain

# ---- LangChain Core Packages ----
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain_core.documents import Document  # ✅ Correct import

# ---- Text Splitters ----
from langchain_text_splitters import RecursiveCharacterTextSplitter

# ---- LLM ----
from langchain_groq import ChatGroq



st.title("RockyBot: News Research Tool 📈")
st.sidebar.title("News Article URLs")


# Collect URLs from user input
urls = [st.sidebar.text_input(f"URL {i+1}") for i in range(3)]
process_url_clicked = st.sidebar.button("Process URLs")
file_path = "faiss_store_openai.pkl"


main_placeholder = st.empty()
llm = ChatGroq(
    api_key=os.environ["GROQ_API_KEY"],
    model_name="llama-3.1-8b-instant"
)


def fetch_web_content(url):
    """Fetch text content from URL using BeautifulSoup."""
    try:
        response = requests.get(url, timeout=10)
        response.raise_for_status()
        soup = BeautifulSoup(response.text, "html.parser")
        return soup.get_text()
    except Exception as e:
        return f"Error fetching {url}: {str(e)}"


if process_url_clicked:
    main_placeholder.text("Data Loading...Started...✅")

    data = [(url, fetch_web_content(url)) for url in urls if url.strip()]

    main_placeholder.text("Data Loading...Completed...✅")

    # Split into chunks
    text_splitter = RecursiveCharacterTextSplitter(
        separators=['\n\n', '\n', '.', ','],
        chunk_size=1000
    )
    main_placeholder.text("Text Splitting...Started...")

    docs = []
    for url, text in data:
        split_docs = text_splitter.split_text(text)
        docs.extend([Document(page_content=chunk, metadata={"source": url}) for chunk in split_docs])

    main_placeholder.text("Text Splitting...Completed...")

    # Embeddings + FAISS
    embedding_model = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
    vectorstore = FAISS.from_documents(docs, embedding_model)

    with open(file_path, "wb") as f:
        pickle.dump(vectorstore, f)

    main_placeholder.text("Vectorstore Ready! ✅")


# --------------------------
# 🔥 NEW CHAIN (REPLACES RetrievalQAWithSourcesChain)
# --------------------------

prompt = ChatPromptTemplate.from_template("""
You are a financial news analyst.

Use ONLY the provided context to answer.

Context:
{context}

Question: {question}

Provide:
- a concise answer  
- list of sources at the end
""")

def build_qa_with_sources_chain(retriever):
    """
    New-style LangChain retrieval chain that returns answer + sources.
    """

    # LLM + prompt → document chain
    document_chain = create_stuff_documents_chain(
        llm=llm,
        prompt=prompt,
        output_parser=StrOutputParser()
    )

    # Retriever + doc chain
    retrieval_chain = create_retrieval_chain(
        retriever=retriever,
        combine_docs_chain=document_chain,
    )

    # Wrapper to format consistent outputs
    class QAWrapper:
        def __init__(self, chain):
            self.chain = chain

        def invoke(self, query):
            result = self.chain.invoke({"question": query})

            answer = result.get("answer") or result.get("result") or ""
            docs = result.get("context") or []

            sources = [d.metadata.get("source", "N/A") for d in docs]

            return {"answer": answer, "sources": sources}

    return QAWrapper(retrieval_chain)


# --------------------------
# QUERY EXECUTION
# --------------------------

query = st.text_input("Question: ")

if query:
    if os.path.exists(file_path):
        # Load FAISS DB
        with open(file_path, "rb") as f:
            vectorstore = pickle.load(f)

        retriever = vectorstore.as_retriever(search_kwargs={"k": 4})

        qa_chain = build_qa_with_sources_chain(retriever)
        result = qa_chain.invoke(query)

        # Display result
        st.header("Answer")
        st.write(result["answer"])

        st.subheader("Sources")
        if result["sources"]:
            for s in result["sources"]:
                st.write("🔗", s)
        else:
            st.write("No sources found.")