Create app.py

app.py

@@ -0,0 +1,159 @@
+import streamlit as st
+import torch
+from transformers import GPTNeoXForCausalLM, AutoTokenizer
+from sentence_transformers import SentenceTransformer
+import faiss
+import fitz  # PyMuPDF
+from langchain_text_splitters import RecursiveCharacterTextSplitter
+
+# Configuration
+MODEL_NAME = "ibm-granite/granite-3.1-1b-a400m-instruct"
+EMBED_MODEL = "sentence-transformers/all-mpnet-base-v2"
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+CHUNK_SIZE = 512
+CHUNK_OVERLAP = 50
+
+@st.cache_resource
+def load_models():
+    try:
+        # Load Granite model
+        tokenizer = AutoTokenizer.from_pretrained(
+            MODEL_NAME,
+            trust_remote_code=True
+        )
+        
+        model = GPTNeoXForCausalLM.from_pretrained(
+            MODEL_NAME,
+            device_map="auto" if DEVICE == "cuda" else None,
+            torch_dtype=torch.float16 if DEVICE == "cuda" else torch.float32,
+            trust_remote_code=True
+        ).eval()
+        
+        # Load sentence transformer for embeddings
+        embedder = SentenceTransformer(EMBED_MODEL, device=DEVICE)
+        
+        return tokenizer, model, embedder
+    
+    except Exception as e:
+        st.error(f"Model loading failed: {str(e)}")
+        st.stop()
+
+tokenizer, model, embedder = load_models()
+
+# Text processing
+def process_text(text):
+    splitter = RecursiveCharacterTextSplitter(
+        chunk_size=CHUNK_SIZE,
+        chunk_overlap=CHUNK_OVERLAP,
+        length_function=len
+    )
+    return splitter.split_text(text)
+
+# PDF extraction
+def extract_pdf_text(uploaded_file):
+    try:
+        doc = fitz.open(stream=uploaded_file.read(), filetype="pdf")
+        return "\n".join([page.get_text() for page in doc])
+    except Exception as e:
+        st.error(f"PDF extraction error: {str(e)}")
+        return ""
+
+# Summarization function
+def generate_summary(text):
+    chunks = process_text(text)[:10]
+    summaries = []
+    
+    for chunk in chunks:
+        prompt = f"""<|user|>
+        Summarize this text section focusing on key themes, characters, and plot points:
+        {chunk[:2000]}
+        <|assistant|>
+        """
+        
+        inputs = tokenizer(prompt, return_tensors="pt").to(DEVICE)
+        outputs = model.generate(**inputs, max_new_tokens=300, temperature=0.3)
+        summaries.append(tokenizer.decode(outputs[0], skip_special_tokens=True))
+    
+    combined = "\n".join(summaries)
+    final_prompt = f"""<|user|>
+    Combine these section summaries into a coherent book summary:
+    {combined}
+    <|assistant|>
+    The comprehensive summary is:"""
+    
+    inputs = tokenizer(final_prompt, return_tensors="pt").to(DEVICE)
+    outputs = model.generate(**inputs, max_new_tokens=500, temperature=0.5)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True).split(":")[-1].strip()
+
+# FAISS index creation
+def build_faiss_index(texts):
+    embeddings = embedder.encode(texts, show_progress_bar=False)
+    dimension = embeddings.shape[1]
+    index = faiss.IndexFlatIP(dimension)
+    faiss.normalize_L2(embeddings)
+    index.add(embeddings)
+    return index
+
+# Answer generation
+def generate_answer(query, context):
+    prompt = f"""<|user|>
+    Using this context: {context}
+    Answer the question precisely and truthfully. If unsure, say "I don't know".
+    Question: {query}
+    <|assistant|>
+    """
+    
+    inputs = tokenizer(prompt, return_tensors="pt", max_length=1024, truncation=True).to(DEVICE)
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=300,
+        temperature=0.4,
+        top_p=0.9,
+        repetition_penalty=1.2,
+        do_sample=True
+    )
+    return tokenizer.decode(outputs[0], skip_special_tokens=True).split("<|assistant|>")[-1].strip()
+
+# Streamlit UI
+st.set_page_config(page_title="📚 Smart Book Analyst", layout="wide")
+st.title("📚 AI-Powered Book Analysis System")
+
+uploaded_file = st.file_uploader("Upload book (PDF or TXT)", type=["pdf", "txt"])
+
+if uploaded_file:
+    with st.spinner("📖 Analyzing book content..."):
+        try:
+            if uploaded_file.type == "application/pdf":
+                text = extract_pdf_text(uploaded_file)
+            else:
+                text = uploaded_file.read().decode()
+            
+            chunks = process_text(text)
+            st.session_state.docs = chunks
+            st.session_state.index = build_faiss_index(chunks)
+            
+            with st.expander("📝 Book Summary", expanded=True):
+                summary = generate_summary(text)
+                st.write(summary)
+                
+        except Exception as e:
+            st.error(f"Processing failed: {str(e)}")
+
+if 'index' in st.session_state and st.session_state.index:
+    query = st.text_input("Ask about the book:")
+    if query:
+        with st.spinner("🔍 Searching for answers..."):
+            try:
+                query_embed = embedder.encode([query])
+                faiss.normalize_L2(query_embed)
+                distances, indices = st.session_state.index.search(query_embed, k=3)
+                
+                context = "\n".join([st.session_state.docs[i] for i in indices[0]])
+                answer = generate_answer(query, context)
+                
+                st.subheader("Answer")
+                st.markdown(f"```\n{answer}\n```")
+                st.caption("Retrieved context confidence: {:.2f}".format(distances[0][0]))
+                
+            except Exception as e:
+                st.error(f"Query failed: {str(e)}")