updated app.py for the input handling

app.py

@@ -67,14 +67,61 @@ def ensure_complete_sentences(text):
         return ' '.join(s.strip() for s in sentences)
     return text
 
+# --- Added: Handling "Not Feasible" Keywords and Gibberish Inputs ---
 def is_valid_input(text):
+    """
+    Validate the user's input question.
+    Returns a tuple (is_valid, message).
+    """
     if not text or text.strip() == "":
-        return False
+        return False, "Input cannot be empty. Please provide a meaningful question."
+    
     if not re.search('[A-Za-z]', text):
-        return False
+        return False, "Input must contain alphabetic characters."
+    
     if len(text.strip()) < 5:
-        return False
-    return True
+        return False, "Input is too short. Please provide a more detailed question."
+    
+    # Define not feasible keywords
+    not_feasible_keywords = [
+        "illegal", "harmful", "dangerous", "unethical", "inappropriate",
+        "forbidden", "restricted", "banned", "prohibited", "secret"
+    ]
+    
+    # Check for not feasible keywords (case-insensitive)
+    pattern = re.compile(r'\b(' + '|'.join(not_feasible_keywords) + r')\b', re.IGNORECASE)
+    if pattern.search(text):
+        return False, "Your question contains restricted or inappropriate content. Please modify your query."
+    
+    # --- Added: Gibberish Detection ---
+    # Simple heuristic: Check the ratio of alphabetic characters to total characters
+    total_chars = len(text)
+    alpha_chars = len(re.findall(r'[A-Za-z]', text))
+    ratio = alpha_chars / total_chars if total_chars > 0 else 0
+    
+    if ratio < 0.6:
+        return False, "Your input appears to be gibberish or nonsensical. Please enter a clear and meaningful question."
+    
+    # Additionally, check for a minimum number of recognizable words
+    words = re.findall(r'\b\w+\b', text)
+    recognized_words = [word for word in words if word.lower() in recognized_words_set]
+    
+    if len(recognized_words) < max(3, len(words) * 0.4):
+        return False, "Your input contains too many unrecognizable words. Please enter a clear and meaningful question."
+    
+    return True, "Valid input."
+
+# Predefined set of common English words for basic gibberish detection
+# In a production environment, consider using a more comprehensive dictionary or language model
+recognized_words_set = set([
+    'the', 'be', 'to', 'of', 'and', 'a', 'in', 'that', 'have', 'I',
+    'it', 'for', 'not', 'on', 'with', 'he', 'as', 'you', 'do', 'at',
+    'this', 'but', 'his', 'by', 'from', 'they', 'we', 'say', 'her',
+    'she', 'or', 'an', 'will', 'my', 'one', 'all', 'would', 'there',
+    'their', 'what', 'so', 'up', 'out', 'if', 'about', 'who', 'get',
+    'which', 'go', 'me'
+    # Add more words as needed
+])
 
 def initialize_llm(model, temperature, max_tokens):
     prompt_allocation = int(max_tokens * 0.2)
@@ -109,10 +156,16 @@ def create_rag_pipeline(file_paths, model, temperature, max_tokens):
 
     retriever = vectorstore.as_retriever()
 
+    # --- Improved Prompt Template ---
     custom_prompt_template = PromptTemplate(
         input_variables=["context", "question"],
         template="""
 You are an AI assistant specialized in daily wellness. Provide a concise, thorough, and stand-alone answer to the user's question based on the given context. Include relevant examples or schedules where beneficial. **When listing steps or guidelines, format them as a numbered list with appropriate markdown formatting.** The final answer should be coherent, self-contained, and end with a complete sentence.
+
+If the question contains restricted or inappropriate content, respond with a polite message indicating that you cannot assist with that request.
+
+If the question appears to be gibberish or nonsensical, respond with a polite message requesting clarification or a more coherent question.
+
 Context:
 {context}
 Question:
@@ -136,8 +189,9 @@ max_tokens = 500
 rag_chain, message = create_rag_pipeline(file_paths, model, temperature, max_tokens)
 
 def answer_question(model, temperature, max_tokens, question):
-    if not is_valid_input(question):
-        return "Please provide a valid, meaningful question."
+    is_valid, message = is_valid_input(question)
+    if not is_valid:
+        return message
     if rag_chain is None:
         return "The system is currently unavailable. Please try again later."
     try: