Update app.py

app.py

@@ -1,203 +1,104 @@
 import os
-import re
-import tempfile
 import requests
-import streamlit as st
-
-# LangChain
-from langchain.text_splitter import RecursiveCharacterTextSplitter
-from langchain.vectorstores import FAISS
-from langchain.schema import Document
-from langchain_groq import ChatGroq
+from groq import Groq
 from langchain_community.embeddings import HuggingFaceEmbeddings
-
-# PDF/DOCX/OCR
+from langchain_community.vectorstores import FAISS
+from langchain.text_splitter import RecursiveCharacterTextSplitter
 from PyPDF2 import PdfReader
-import docx
-from pdf2image import convert_from_path
-import pytesseract
-
-# ==============================
-# CONFIG
-# ==============================
-
-st.set_page_config(page_title="Robust Google Drive RAG", layout="wide")
-
-GROQ_MODEL_NAME = "llama-3.1-8b-instant"
-GROQ_API_KEY = os.getenv("API")
-if not GROQ_API_KEY:
-    st.error("GROQ_API_KEY not set. Add it in HF Space → Settings → Secrets.")
-    st.stop()
-
-EMBED_MODEL = "sentence-transformers/all-MiniLM-L6-v2"
-CHUNK_SIZE = 1000
-CHUNK_OVERLAP = 200
-TOP_K = 4
-
-# ==============================
-# GOOGLE DRIVE
-# ==============================
-
-def extract_drive_file_id(url: str) -> str:
-    match = re.search(r"/d/([a-zA-Z0-9_-]+)", url)
-    return match.group(1) if match else None
-
-def download_drive_file(file_id: str, destination: str):
-    url = "https://drive.google.com/uc?export=download"
-    session = requests.Session()
-    response = session.get(url, params={"id": file_id}, stream=True)
-    for k, v in response.cookies.items():
-        if k.startswith("download_warning"):
-            response = session.get(url, params={"id": file_id, "confirm": v}, stream=True)
-    if response.status_code != 200:
-        raise RuntimeError("Failed to download file. Check link permissions.")
-    with open(destination, "wb") as f:
-        for chunk in response.iter_content(32768):
-            f.write(chunk)
-
-# ==============================
-# LOAD DOCUMENT
-# ==============================
+import streamlit as st
+from tempfile import NamedTemporaryFile
 
-def load_pdf(path):
-    reader = PdfReader(path)
-    text = "\n".join(page.extract_text() for page in reader.pages if page.extract_text())
-    return text
+# Initialize Groq client
+client = Groq(api_key=os.environ['API'])
 
-def load_pdf_ocr(path):
-    pages = convert_from_path(path)
+# Function to extract text from a PDF
+def extract_text_from_pdf(pdf_file_path):
+    pdf_reader = PdfReader(pdf_file_path)
     text = ""
-    for page in pages:
-        text += pytesseract.image_to_string(page)
+    for page in pdf_reader.pages:
+        page_text = page.extract_text()
+        if page_text:
+            text += page_text
     return text
 
-def load_docx(path):
-    doc = docx.Document(path)
-    return "\n".join(p.text for p in doc.paragraphs if p.text.strip())
-
-def load_txt(path):
-    with open(path, "r", encoding="utf-8", errors="ignore") as f:
-        return f.read()
-
-def load_document(path):
-    ext = path.split(".")[-1].lower()
-    text = ""
-    if ext == "pdf":
-        text = load_pdf(path)
-        if not text.strip():  # fallback to OCR
-            text = load_pdf_ocr(path)
-    elif ext == "docx":
-        text = load_docx(path)
-    elif ext == "txt":
-        text = load_txt(path)
-    else:
-        raise ValueError("Unsupported file type. Only PDF/DOCX/TXT allowed.")
-    return text.replace("\xa0", " ").replace("\n\n", "\n").strip()
-
-# ==============================
-# EMBEDDINGS
-# ==============================
-
-@st.cache_resource
-def load_embeddings():
-    return HuggingFaceEmbeddings(model_name=EMBED_MODEL)
-
-# ==============================
-# VECTOR STORE
-# ==============================
-
-def build_vectorstore(text: str):
-    splitter = RecursiveCharacterTextSplitter(
-        chunk_size=CHUNK_SIZE,
-        chunk_overlap=CHUNK_OVERLAP
+# Function to split text into chunks
+def chunk_text(text, chunk_size=500, chunk_overlap=50):
+    text_splitter = RecursiveCharacterTextSplitter(
+        chunk_size=chunk_size, chunk_overlap=chunk_overlap
     )
-    chunks = splitter.split_text(text)
-    docs = [Document(page_content=c) for c in chunks]
-    vs = FAISS.from_documents(docs, load_embeddings())
-    return vs, len(chunks)
+    return text_splitter.split_text(text)
 
-# ==============================
-# QA
-# ==============================
-
-def answer_question(question, vectorstore):
-    docs = vectorstore.similarity_search(question, k=TOP_K)
-    if not docs:
-        return "The answer is not available in the provided document.", []
-
-    context = "\n\n".join(d.page_content for d in docs)
-
-    system_prompt = (
-        "You must answer ONLY from the provided context.\n"
-        "If the answer is missing, reply exactly:\n"
-        "'The answer is not available in the provided document.'\n"
-        "No external knowledge allowed."
-    )
-
-    llm = ChatGroq(
-        api_key=GROQ_API_KEY,
-        model_name=GROQ_MODEL_NAME,
-        temperature=0
+# Function to create embeddings and store them in FAISS
+def create_embeddings_and_store(chunks, vector_db=None):
+    embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
+    if vector_db is None:
+        vector_db = FAISS.from_texts(chunks, embedding=embeddings)
+    else:
+        vector_db.add_texts(chunks)
+    return vector_db
+
+# Function to query the vector database and interact with Groq
+def query_vector_db(query, vector_db):
+    docs = vector_db.similarity_search(query, k=3)
+    context = "\n".join([doc.page_content for doc in docs])
+    chat_completion = client.chat.completions.create(
+        messages=[
+            {"role": "system", "content": f"Use the following context:\n{context}"},
+            {"role": "user", "content": query},
+        ],
+        model="llama3-8b-8192",
     )
-
-    response = llm.invoke([
-        {"role": "system", "content": system_prompt},
-        {"role": "user", "content": f"Context:\n{context}\n\nQuestion:\n{question}"}
-    ])
-
-    return response.content.strip(), docs
-
-# ==============================
-# STREAMLIT UI
-# ==============================
-
-st.title("📄 Robust Google Drive RAG App")
-st.caption("Answers strictly from your document. Supports scanned PDFs via OCR.")
-
-if "vectorstore" not in st.session_state:
-    st.session_state.vectorstore = None
-
-with st.sidebar:
-    st.header("📂 Document")
-    drive_link = st.text_input("Google Drive Link")
-
-    if st.button("Process Document"):
-        try:
-            file_id = extract_drive_file_id(drive_link)
-            if not file_id:
-                raise ValueError("Invalid Google Drive link.")
-
-            with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp:
-                download_drive_file(file_id, tmp.name)
-                text = load_document(tmp.name)
-
-            if not text.strip():
-                raise ValueError("Document appears empty.")
-
-            vectorstore, chunks = build_vectorstore(text)
-            st.session_state.vectorstore = vectorstore
-            st.success(f"Document indexed successfully ({chunks} chunks).")
-
-        except Exception as e:
-            st.error(str(e))
-
-    if st.button("Reset"):
-        st.session_state.vectorstore = None
-        st.success("Vector store cleared.")
-
-st.subheader("❓ Ask a Question")
-question = st.text_input("Your question")
-
-if st.button("Get Answer"):
-    if not st.session_state.vectorstore:
-        st.error("Process a document first.")
+    return chat_completion.choices[0].message.content
+
+# Function to convert Google Drive view link to downloadable link
+def get_direct_download_link(view_url):
+    if "drive.google.com/file/d/" in view_url:
+        file_id = view_url.split("/file/d/")[1].split("/")[0]
+        return f"https://drive.google.com/uc?export=download&id={file_id}"
+    return None
+
+# Function to download and save a PDF from a URL
+def download_pdf_from_url(url):
+    direct_url = get_direct_download_link(url)
+    if not direct_url:
+        return None
+    response = requests.get(direct_url)
+    if response.status_code == 200:
+        temp_file = NamedTemporaryFile(delete=False, suffix=".pdf")
+        temp_file.write(response.content)
+        temp_file.close()
+        return temp_file.name
     else:
-        answer, docs = answer_question(question, st.session_state.vectorstore)
-        st.markdown("### ✅ Answer")
-        st.write(answer)
-
-        with st.expander("🔍 Retrieved Context"):
-            for i, d in enumerate(docs, 1):
-                st.markdown(f"**Chunk {i}**")
-                st.write(d.page_content)
+        return None
+
+# Streamlit app
+st.title("RAG-Based QA on Google Drive PDFs")
+
+# Only fetch from provided links
+doc_links = [
+    "https://docs.google.com/document/d/1_3DqDH0WVFQcwb23EwFWVfHCC0fCv58p/edit?usp=drive_link&ouid=108024737681201137204&rtpof=true&sd=true",
+]
+
+vector_db = None
+
+# Process Google Drive documents
+for idx, link in enumerate(doc_links):
+    st.write(f"📄 Fetching and processing PDF from Link {idx + 1}...")
+    pdf_path = download_pdf_from_url(link)
+    if pdf_path:
+        text = extract_text_from_pdf(pdf_path)
+        chunks = chunk_text(text)
+        vector_db = create_embeddings_and_store(chunks, vector_db=vector_db)
+        st.success(f"✅ Processed document {idx + 1}")
+    else:
+        st.error(f"❌ Failed to download or process PDF from Link {idx + 1}")
+
+# User query input
+user_query = st.text_input("🔍 Enter your query:")
+if user_query and vector_db:
+    response = query_vector_db(user_query, vector_db)
+    st.subheader("💬 Response from LLM:")
+    st.write(response)
+elif user_query:
+    st.warning("⚠️ No documents processed to query.")