Make offline chat model current date aware. Improve system prompts

- Can now expect date awareness chat quality test to pass
- Prevent offline chat model from printing verbatim user Notes and
  special tokens
- Make it ask follow-up questions if it needs more context

src/khoj/processor/conversation/offline/chat_model.py

@@ -171,9 +171,10 @@ def converse_offline(
         )
 
     # Setup Prompt with Primer or Conversation History
+    current_date = datetime.now().strftime("%Y-%m-%d")
     messages = generate_chatml_messages_with_context(
         conversation_primer,
-        prompts.system_prompt_message_gpt4all,
+        prompts.system_prompt_message_gpt4all.format(current_date=current_date),
         conversation_log,
         model_name=model,
         max_prompt_size=max_prompt_size,
@@ -198,7 +199,7 @@ def llm_thread(g, messages: List[ChatMessage], model: Any):
         for message in conversation_history
     ]
 
-    stop_words = ["<s>"]
+    stop_words = ["<s>", "INST]", "Notes:"]
     chat_history = "".join(formatted_messages)
     templated_system_message = prompts.system_prompt_gpt4all.format(message=system_message.content)
     templated_user_message = prompts.user_message_gpt4all.format(message=user_message.content)

src/khoj/processor/conversation/prompts.py

@@ -5,7 +5,7 @@ from langchain.prompts import PromptTemplate
 personality = PromptTemplate.from_template(
     """
 You are Khoj, a smart, inquisitive and helpful personal assistant.
-Use your general knowledge and the past conversation with the user as context to inform your responses.
+Use your general knowledge and past conversation with the user as context to inform your responses.
 You were created by Khoj Inc. with the following capabilities:
 
 - You *CAN REMEMBER ALL NOTES and PERSONAL INFORMATION FOREVER* that the user ever shares with you.
@@ -48,9 +48,17 @@ no_entries_found = PromptTemplate.from_template(
 
 ## Conversation Prompts for GPT4All Models
 ## --
-system_prompt_message_gpt4all = f"""You are Khoj, a smart, inquisitive and helpful personal assistant.
-Using your general knowledge and our past conversations as context, answer the following question.
-If you do not know the answer, say 'I don't know.'"""
+system_prompt_message_gpt4all = PromptTemplate.from_template(
+    """
+You are Khoj, a smart, inquisitive and helpful personal assistant.
+- Use your general knowledge and past conversation with the user as context to inform your responses.
+- If you do not know the answer, say 'I don't know.'
+- Ask crisp follow-up questions to get additional context, when the answer cannot be inferred from the provided notes or past conversations.
+- Do not print verbatim Notes unless necessary.
+
+Today is {current_date} in UTC.
+    """.strip()
+)
 
 system_prompt_message_extract_questions_gpt4all = f"""You are Khoj, a kind and intelligent personal assistant. When the user asks you a question, you ask follow-up questions to clarify the necessary information you need in order to answer from the user's perspective.
 - Write the question as if you can search for the answer on the user's personal notes.

tests/test_gpt4all_chat_director.py

@@ -298,7 +298,6 @@ def test_answer_not_known_using_notes_command(client_offline_chat, default_user2
 
 
 # ----------------------------------------------------------------------------------------------------
-@pytest.mark.xfail(AssertionError, reason="Chat director not capable of answering time aware questions yet")
 @pytest.mark.chatquality
 @pytest.mark.django_db(transaction=True)
 @freeze_time("2023-04-01", ignore=["transformers"])
@@ -336,7 +335,6 @@ def test_answer_requires_date_aware_aggregation_across_provided_notes(client_off
 
 
 # ----------------------------------------------------------------------------------------------------
-@pytest.mark.xfail(AssertionError, reason="Chat director not capable of answering this question yet")
 @pytest.mark.chatquality
 @pytest.mark.django_db(transaction=True)
 def test_answer_general_question_not_in_chat_history_or_retrieved_content(client_offline_chat, default_user2):