Update app.py

app.py

@@ -11,16 +11,16 @@ from PIL import Image
 
 logging.basicConfig(level=logging.ERROR)
 device = -1  # CPU-only
-print("⚠️ CPU-only. Expect ~20–30s for 300,000 chars.")
+print("⚠️ CPU-only. Using faster models!")
 
-# Load summarizer
+# Load faster summarizer
 try:
-    summarizer = pipeline("summarization", model="t5-small", device=device, torch_dtype=torch.float32)
+    summarizer = pipeline("summarization", model="sshleifer/distilbart-cnn-12-6", device=device, torch_dtype=torch.float32)
 except Exception as e:
     print(f"❌ Summarizer model loading failed: {str(e)}")
     exit(1)
 
-# Load better QA model
+# Load QA model
 try:
     qa_pipeline = pipeline("question-answering", model="deepset/roberta-base-squad2", device=device)
 except Exception as e:
@@ -62,8 +62,8 @@ def summarize_file(file_bytes):
     if not text.strip():
         return "❌ No text found", None
 
-    text = text[:300000]
-    chunks = [text[i:i+2000] for i in range(0, len(text), 2000)][:3]  # Limit to 3 chunks for testing
+    text = text[:200000]  # reduced slightly for speed
+    chunks = [text[i:i+1000] for i in range(0, len(text), 1000)][:5]  # smaller + more chunks
     summaries = []
 
     for i, chunk in enumerate(chunks):
@@ -79,7 +79,7 @@ def summarize_file(file_bytes):
             chunk_result['status'] = 'skipped'
         else:
             try:
-                summary = summarizer(chunk, max_length=60, min_length=10, do_sample=False)[0]['summary_text']
+                summary = summarizer(chunk, max_length=50, min_length=10, do_sample=False)[0]['summary_text']
                 summaries.append(f"### Chunk {i+1}\n{summary}")
                 chunk_result['status'] = 'summarized'
             except Exception as e:
@@ -92,21 +92,29 @@ def summarize_file(file_bytes):
     formatted_chunks = "\n\n---\n\n".join(summaries)
     final_summary = f"""**Characters Processed**: {len(text)}
 **Total Time**: {time.time()-start:.2f} seconds
-
 ## 🔹 Summary by Chunks
-
 {formatted_chunks}
 """
     image = visualize_chunk_status(chunk_info)
     return final_summary, image
 
+def find_relevant_passages(text, question, num_passages=3):
+    passages = re.split(r'(?<=[.?!])\s+', text)
+    scored = []
+    for passage in passages:
+        score = sum(1 for word in question.lower().split() if word in passage.lower())
+        scored.append((score, passage))
+    scored.sort(reverse=True)
+    best_passages = " ".join([p for _, p in scored[:num_passages]])
+    return best_passages
+
 def answer_question(file_bytes, question):
     try:
         doc = fitz.open(stream=file_bytes, filetype="pdf")
         text = "".join(page.get_text("text") for page in doc)
         text = re.sub(r"\s+", " ", text).strip()
         text = "".join(c for c in text if ord(c) < 128)
-        context = text[:300000]
+        context = text[:200000]
     except Exception as e:
         return f"❌ Text extraction failed: {str(e)}"
 
@@ -114,7 +122,8 @@ def answer_question(file_bytes, question):
         return "⚠️ Please enter a valid question."
 
     try:
-        result = qa_pipeline(question=question, context=context)
+        relevant_context = find_relevant_passages(context, question)
+        result = qa_pipeline(question=question, context=relevant_context)
         return f"**Answer**: {result['answer']}\n\n**Score**: {result['score']:.2f}"
     except Exception as e:
         return f"❌ QA failed: {str(e)}"
@@ -128,7 +137,7 @@ summarizer_ui = gr.Interface(
         gr.Image(label="📊 Visual Process Flow", type="pil")
     ],
     title="📝 AI-Powered PDF Summarizer",
-    description="Summarizes long PDFs and visualizes chunk-level processing (limited to 3 chunks for testing)."
+    description="Summarizes long PDFs and visualizes chunk-level processing."
 )
 
 # Q&A UI