Update app.py

app.py

@@ -349,8 +349,8 @@ def ask_question(question, temperature, top_p, repetition_penalty, web_search, c
         database = None
 
     max_attempts = 5
-    context_reduction_factor = 0.5  # More aggressive reduction
-    max_estimated_tokens = 25000  # Further reduced to leave more room for response
+    context_reduction_factor = 0.5
+    max_estimated_tokens = 25000
 
     if web_search:
         contextualized_question, topics, entity_tracker, instructions = chatbot.process_question(question)
@@ -361,7 +361,7 @@ def ask_question(question, temperature, top_p, repetition_penalty, web_search, c
 
         for attempt in range(max_attempts):
             try:
-                web_docs = [Document(page_content=result["text"][:1000], metadata={"source": result["link"]}) for result in search_results if result["text"]]  # Limit each result to 1000 characters
+                web_docs = [Document(page_content=result["text"][:1000], metadata={"source": result["link"]}) for result in search_results if result["text"]]
 
                 if database is None:
                     database = FAISS.from_documents(web_docs, embed)
@@ -392,11 +392,11 @@ def ask_question(question, temperature, top_p, repetition_penalty, web_search, c
 
                 while True:
                     formatted_prompt = prompt_val.format(
-                        context=current_context[:3000],  # Limit context to 3000 characters
-                        conv_context=current_conv_context[:500],  # Limit conversation context to 500 characters
+                        context=current_context[:3000],
+                        conv_context=current_conv_context[:500],
                         question=question,
-                        topics=", ".join(current_topics[:5]),  # Limit to 5 topics
-                        entities=json.dumps({k: v[:2] for k, v in current_entities.items()})  # Limit to 2 entities per type
+                        topics=", ".join(current_topics[:5]),
+                        entities=json.dumps({k: v[:2] for k, v in current_entities.items()})
                     )
 
                     estimated_tokens = estimate_tokens(formatted_prompt)
@@ -404,7 +404,6 @@ def ask_question(question, temperature, top_p, repetition_penalty, web_search, c
                     if estimated_tokens <= max_estimated_tokens:
                         break
                     
-                    # Reduce context if estimated token count is too high
                     current_context = current_context[:int(len(current_context) * context_reduction_factor)]
                     current_conv_context = current_conv_context[:int(len(current_conv_context) * context_reduction_factor)]
                     current_topics = current_topics[:max(1, int(len(current_topics) * context_reduction_factor))]
@@ -449,23 +448,20 @@ def ask_question(question, temperature, top_p, repetition_penalty, web_search, c
                 context_str = "\n".join([doc.page_content for doc in relevant_docs])
 
                 prompt_template = """
-                Answer the question based on the following context from the PDF document:
-                Context:
-                {context}
+                Answer based on: PDF Context: {context}
                 Question: {question}
                 Provide a summarized and direct answer to the question.
                 """
 
                 while True:
                     prompt_val = ChatPromptTemplate.from_template(prompt_template)
-                    formatted_prompt = prompt_val.format(context=context_str, question=question)
+                    formatted_prompt = prompt_val.format(context=context_str[:3000], question=question)
 
                     estimated_tokens = estimate_tokens(formatted_prompt)
                     
                     if estimated_tokens <= max_estimated_tokens:
                         break
                     
-                    # Reduce context if estimated token count is too high
                     context_str = context_str[:int(len(context_str) * context_reduction_factor)]
 
                     if len(context_str) < 100: