Update app.py

app.py

@@ -69,21 +69,35 @@ def truncate_text(text, max_tokens=1000):
         text = tokenizer.decode(tokens)
     return text
 
+def sliding_window(text, window_size=512, stride=256):
+    tokens = tokenizer.encode(text)
+    windows = []
+    for i in range(0, len(tokens), stride):
+        window = tokens[i:i+window_size]
+        windows.append(tokenizer.decode(window))
+    return windows
+
 def chat_with_context(question: str) -> str:
     if not indexed_texts:
         return "Please index some text first."
     
     context = find_most_similar(question)[0]
-    context = truncate_text(context, max_tokens=800)  # Truncate context
-    question = truncate_text(question, max_tokens=200)  # Truncate question
-    prompt = f"Context: {context}\n\nQuestion: {question}\n\nAnswer:"
+    windows = sliding_window(context, window_size=512, stride=256)
+    question = truncate_text(question, max_tokens=100)  # Further reduce question length
     
-    try:
-        response = chat_model(prompt, max_length=1024, num_return_sequences=1)
-        answer = response[0]['generated_text'].split("Answer:")[-1].strip()
-        return answer
-    except Exception as e:
-        return f"Error in chat: {str(e)}"
+    responses = []
+    for window in windows:
+        prompt = f"Context: {window}\n\nQuestion: {question}\n\nAnswer:"
+        try:
+            response = chat_model(prompt, max_length=1024, num_return_sequences=1)
+            answer = response[0]['generated_text'].split("Answer:")[-1].strip()
+            responses.append(answer)
+        except Exception as e:
+            logging.error(f"Error in chat: {str(e)}")
+    
+    # Combine and truncate responses
+    combined_response = " ".join(responses)
+    return truncate_text(combined_response, max_tokens=500)
 
 # Translation function
 def translate_text(text, target_lang_code):