Update app.py

app.py

@@ -7,6 +7,7 @@ from sentence_transformers import SentenceTransformer
 import faiss
 import numpy as np
 from groq import Groq
+import docx  # to read .docx files
 
 # --- Helper Functions ---
 
@@ -25,6 +26,17 @@ def extract_text_from_pdf(pdf_path):
         st.warning(f"PyPDF2 failed with error: {e}. Trying pdfminer.six...")
         return extract_text(pdf_path)
 
+def extract_text_from_docx(docx_path):
+    try:
+        doc = docx.Document(docx_path)
+        full_text = []
+        for para in doc.paragraphs:
+            full_text.append(para.text)
+        return '\n'.join(full_text)
+    except Exception as e:
+        st.warning(f"Failed to read DOCX {docx_path}: {e}")
+        return ""
+
 def chunk_text_with_tokenizer(text, tokenizer, chunk_size=150, chunk_overlap=30):
     tokens = tokenizer.tokenize(text)
     chunks = []
@@ -51,9 +63,9 @@ def generate_answer_with_groq(question, context):
         response = groq_client.chat.completions.create(
             model=model_name,
             messages=[
-            {"role": "system", "content": "You are an AI Assistant for Small Businesses.You are an SME expert."},
-            {"role": "user", "content": prompt},
-        ]
+                {"role": "system", "content": "You are an AI Assistant for Small Businesses. You are an SME expert."},
+                {"role": "user", "content": prompt},
+            ]
         )
         return response.choices[0].message.content
     except Exception as e:
@@ -73,58 +85,62 @@ if not GROQ_API_KEY:
 
 os.environ["GROQ_API_KEY"] = GROQ_API_KEY
 
-# File uploader
-uploaded_pdf = st.file_uploader("📁 Upload PDF document(s) for SME knowledge base", type=["pdf"], accept_multiple_files=False)
+# Load and process all docs at startup
+@st.cache_data(show_spinner=True)
+def load_and_prepare_docs(folder_path="docs"):
+    all_text = ""
+    if not os.path.exists(folder_path):
+        st.error(f"Folder '{folder_path}' does not exist!")
+        return None, None, None
+
+    # Collect all pdf and docx files
+    files = [f for f in os.listdir(folder_path) if f.lower().endswith(('.pdf', '.docx', '.doc'))]
+    if not files:
+        st.error(f"No PDF or DOCX files found in folder '{folder_path}'.")
+        return None, None, None
+
+    for file in files:
+        path = os.path.join(folder_path, file)
+        if file.lower().endswith('.pdf'):
+            text = extract_text_from_pdf(path)
+        elif file.lower().endswith(('.docx', '.doc')):
+            text = extract_text_from_docx(path)
+        else:
+            text = ""
+        if text:
+            all_text += text + "\n\n"
+
+    if not all_text.strip():
+        st.error("No text extracted from documents.")
+        return None, None, None
+
+    tokenizer = AutoTokenizer.from_pretrained('bert-base-uncased')
+    text_chunks = chunk_text_with_tokenizer(all_text, tokenizer)
+
+    embedding_model = SentenceTransformer('all-mpnet-base-v2')
+    all_embeddings = embedding_model.encode(text_chunks) if text_chunks else None
+
+    if all_embeddings is None or len(all_embeddings) == 0:
+        st.error("No text chunks found to create embeddings.")
+        return None, None, None
+
+    embedding_dim = all_embeddings[0].shape[0]
+    index = faiss.IndexFlatL2(embedding_dim)
+    index.add(np.array(all_embeddings))
+
+    return index, embedding_model, text_chunks
+
+index, embedding_model, text_chunks = load_and_prepare_docs()
 
-# Text input for question
 user_question = st.text_input("💬 Ask your question about SME documents:")
 
-# Button to trigger processing
-if st.button("Get Answer") or (user_question and uploaded_pdf):
-    if not uploaded_pdf:
-        st.warning("Please upload a PDF file first.")
-    elif not user_question:
-        st.warning("Please enter a question.")
+if st.button("Get Answer") and user_question:
+    if index is None or embedding_model is None or text_chunks is None:
+        st.error("The document knowledge base is not ready. Please check the errors above.")
     else:
-        with st.spinner("Processing PDF and generating answer..."):
-            # Save uploaded file temporarily for PyPDF2/pdfminer
-            temp_path = f"/tmp/{uploaded_pdf.name}"
-            with open(temp_path, "wb") as f:
-                f.write(uploaded_pdf.getbuffer())
-
-            # Extract text
-            pdf_text = extract_text_from_pdf(temp_path)
-
-            # Tokenizer + Chunk
-            tokenizer = AutoTokenizer.from_pretrained('bert-base-uncased')
-            text_chunks = chunk_text_with_tokenizer(pdf_text, tokenizer)
-
-            # Embeddings
-            embedding_model = SentenceTransformer('all-mpnet-base-v2')
-            all_embeddings = embedding_model.encode(text_chunks) if text_chunks else None
-
-            if all_embeddings is None or len(all_embeddings) == 0:
-                st.error("No text chunks found to create embeddings.")
-            else:
-                # Create FAISS index
-                embedding_dim = all_embeddings[0].shape[0]
-                index = faiss.IndexFlatL2(embedding_dim)
-                index.add(np.array(all_embeddings))
-
-                # Retrieve relevant chunks
-                relevant_chunks = retrieve_relevant_chunks(user_question, index, embedding_model, text_chunks)
-                context = "\n\n".join(relevant_chunks)
-
-                # Generate answer with Groq
-                answer = generate_answer_with_groq(user_question, context)
-
-                # Display outputs
-                #st.markdown("### Extracted Text Snippet:")
-                #st.write(pdf_text[:500] + "...")
-
-                #st.markdown("### Sample Text Chunks:")
-                #for i, chunk in enumerate(text_chunks[:3]):
-                #    st.write(f"Chunk {i+1}: {chunk[:200]}...")
-
-                st.markdown("### Answer:")
-                st.success(answer)
+        with st.spinner("Searching for relevant information and generating answer..."):
+            relevant_chunks = retrieve_relevant_chunks(user_question, index, embedding_model, text_chunks)
+            context = "\n\n".join(relevant_chunks)
+            answer = generate_answer_with_groq(user_question, context)
+            st.markdown("### Answer:")
+            st.success(answer)