revised app.py

app.py

@@ -71,21 +71,22 @@ def load_documents(file_paths):
             logger.error(f"Error processing file {file_path}: {e}")
     return docs
 
-# Function to ensure the response ends with a complete sentence
+# Function to ensure the response ends with complete sentences
 def ensure_complete_sentences(text):
     # Use regex to find all complete sentences
     sentences = re.findall(r'[^.!?]*[.!?]', text)
     if sentences:
-        return sentences[-1].strip()
+        # Join all complete sentences to form the complete answer
+        return ' '.join(sentences).strip()
     return text  # Return as is if no complete sentence is found
 
 # Initialize the LLM using ChatGroq with GROQ's API
 def initialize_llm(model, temperature, max_tokens):
     try:
-        # Allocate some tokens for the prompt (e.g., 50 tokens)
-        prompt_tokens = 50
-        response_max_tokens = max_tokens - prompt_tokens
-        if response_max_tokens <= 0:
+        # Allocate a portion of tokens for the prompt, e.g., 20%
+        prompt_allocation = int(max_tokens * 0.2)
+        response_max_tokens = max_tokens - prompt_allocation
+        if response_max_tokens <= 50:
             raise ValueError("max_tokens is too small to allocate for the response.")
 
         llm = ChatGroq(
@@ -129,7 +130,7 @@ def create_rag_pipeline(file_paths, model, temperature, max_tokens):
         custom_prompt_template = PromptTemplate(
             input_variables=["context", "question"],
             template="""
-            You are an AI assistant with expertise in daily wellness. Your aim is to provide detailed yet concise solutions regarding daily wellness topics.
+            You are an AI assistant with expertise in daily wellness. Your aim is to provide detailed and comprehensive solutions regarding daily wellness topics without unnecessary verbosity.
 
             Context:
             {context}
@@ -137,7 +138,7 @@ def create_rag_pipeline(file_paths, model, temperature, max_tokens):
             Question:
             {question}
 
-            Provide a detailed but concise answer, ensuring that it is complete and does not end abruptly. Include relevant examples and a suggested schedule.
+            Provide a thorough and complete answer, including relevant examples and a suggested schedule. Ensure that the response does not end abruptly.
             """
         )
 
@@ -161,7 +162,7 @@ def answer_question(file_paths, model, temperature, max_tokens, question):
     try:
         answer = rag_chain.run(question)
         logger.debug("Question answered successfully.")
-        # Post-process to ensure the answer ends with a complete sentence
+        # Post-process to ensure the answer ends with complete sentences
         complete_answer = ensure_complete_sentences(answer)
         return complete_answer
     except Exception as e:
@@ -179,7 +180,7 @@ interface = gr.Interface(
     inputs=[
         gr.Textbox(label="Model Name", value="llama3-8b-8192"),
         gr.Slider(label="Temperature", minimum=0, maximum=1, step=0.01, value=0.7),
-        gr.Slider(label="Max Tokens", minimum=100, maximum=1024, step=1, value=500),
+        gr.Slider(label="Max Tokens", minimum=200, maximum=1024, step=1, value=500),
         gr.Textbox(label="Question")
     ],
     outputs="text",