use a consistent warning prefix across codebase (#569)

memgpt/agent.py

@@ -18,6 +18,7 @@
     CORE_MEMORY_HUMAN_CHAR_LIMIT,
     CORE_MEMORY_PERSONA_CHAR_LIMIT,
     LLM_MAX_TOKENS,
+    CLI_WARNING_PREFIX,
 )
 from .errors import LLMError
 from .functions.functions import load_all_function_sets
@@ -505,7 +506,7 @@ def handle_ai_response(self, response_message):
             heartbeat_request = function_args.pop("request_heartbeat", None)
             if not (isinstance(heartbeat_request, bool) or heartbeat_request is None):
                 printd(
-                    f"Warning: 'request_heartbeat' arg parsed was not a bool or None, type={type(heartbeat_request)}, value={heartbeat_request}"
+                    f"{CLI_WARNING_PREFIX}'request_heartbeat' arg parsed was not a bool or None, type={type(heartbeat_request)}, value={heartbeat_request}"
                 )
                 heartbeat_request = None
 
@@ -568,7 +569,7 @@ def step(self, user_message, first_message=False, first_message_retry_limit=FIRS
                 input_message_sequence = self.messages
 
             if len(input_message_sequence) > 1 and input_message_sequence[-1]["role"] != "user":
-                printd(f"WARNING: attempting to run ChatCompletion without user as the last message in the queue")
+                printd(f"{CLI_WARNING_PREFIX}Attempting to run ChatCompletion without user as the last message in the queue")
 
             # Step 1: send the conversation and available functions to GPT
             if not skip_verify and (first_message or self.messages_total == self.messages_total_init):
@@ -620,7 +621,7 @@ def step(self, user_message, first_message=False, first_message_retry_limit=FIRS
             # We can't do summarize logic properly if context_window is undefined
             if self.config.context_window is None:
                 # Fallback if for some reason context_window is missing, just set to the default
-                print(f"WARNING: could not find context_window in config, setting to default {LLM_MAX_TOKENS['DEFAULT']}")
+                print(f"{CLI_WARNING_PREFIX}could not find context_window in config, setting to default {LLM_MAX_TOKENS['DEFAULT']}")
                 print(f"{self.config}")
                 self.config.context_window = (
                     str(LLM_MAX_TOKENS[self.model])
@@ -629,7 +630,7 @@ def step(self, user_message, first_message=False, first_message_retry_limit=FIRS
                 )
             if current_total_tokens > MESSAGE_SUMMARY_WARNING_FRAC * int(self.config.context_window):
                 printd(
-                    f"WARNING: last response total_tokens ({current_total_tokens}) > {MESSAGE_SUMMARY_WARNING_FRAC * int(self.config.context_window)}"
+                    f"{CLI_WARNING_PREFIX}last response total_tokens ({current_total_tokens}) > {MESSAGE_SUMMARY_WARNING_FRAC * int(self.config.context_window)}"
                 )
                 # Only deliver the alert if we haven't already (this period)
                 if not self.agent_alerted_about_memory_pressure:
@@ -712,7 +713,7 @@ def summarize_messages_inplace(self, cutoff=None, preserve_last_N_messages=True)
         # We can't do summarize logic properly if context_window is undefined
         if self.config.context_window is None:
             # Fallback if for some reason context_window is missing, just set to the default
-            print(f"WARNING: could not find context_window in config, setting to default {LLM_MAX_TOKENS['DEFAULT']}")
+            print(f"{CLI_WARNING_PREFIX}could not find context_window in config, setting to default {LLM_MAX_TOKENS['DEFAULT']}")
             print(f"{self.config}")
             self.config.context_window = (
                 str(LLM_MAX_TOKENS[self.model])

memgpt/autogen/interface.py

@@ -3,6 +3,8 @@
 
 from colorama import Fore, Style, init
 
+from memgpt.constants import CLI_WARNING_PREFIX
+
 init(autoreset=True)
 
 
@@ -104,7 +106,7 @@ def user_message(self, msg, raw=False):
                 try:
                     msg_json = json.loads(msg)
                 except:
-                    print(f"Warning: failed to parse user message into json")
+                    print(f"{CLI_WARNING_PREFIX}failed to parse user message into json")
                     message = f"{Fore.GREEN}{Style.BRIGHT}🧑 {Fore.GREEN}{msg}{Style.RESET_ALL}" if self.fancy else f"[user] {msg}"
                     self.message_list.append(message)
                     return
@@ -175,7 +177,7 @@ def function_message(self, msg):
                             print(e)
                             message = msg_dict
                     else:
-                        print(f"Warning: did not recognize function message")
+                        print(f"{CLI_WARNING_PREFIX}did not recognize function message")
                         message = (
                             f"{Fore.RED}{Style.BRIGHT}⚡ [function] {Fore.RED}{msg}{Style.RESET_ALL}" if self.fancy else f"[function] {msg}"
                         )
@@ -194,7 +196,7 @@ def function_message(self, msg):
                         f"{Fore.GREEN}{Style.BRIGHT}⚡ [function] {Fore.GREEN}{msg}{Style.RESET_ALL}" if self.fancy else f"[function] {msg}"
                     )
             except Exception:
-                print(f"Warning: did not recognize function message {type(msg)} {msg}")
+                print(f"{CLI_WARNING_PREFIX}did not recognize function message {type(msg)} {msg}")
                 message = f"{Fore.RED}{Style.BRIGHT}⚡ [function] {Fore.RED}{msg}{Style.RESET_ALL}" if self.fancy else f"[function] {msg}"
 
         if message:

memgpt/cli/cli.py

@@ -15,7 +15,7 @@
 from memgpt.utils import printd
 from memgpt.persistence_manager import LocalStateManager
 from memgpt.config import MemGPTConfig, AgentConfig
-from memgpt.constants import MEMGPT_DIR
+from memgpt.constants import MEMGPT_DIR, CLI_WARNING_PREFIX
 from memgpt.agent import Agent
 from memgpt.embeddings import embedding_model
 
@@ -107,36 +107,39 @@ def run(
         # persistence_manager = LocalStateManager(agent_config).load() # TODO: implement load
         # TODO: load prior agent state
         if persona and persona != agent_config.persona:
-            typer.secho(f"Warning: Overriding existing persona {agent_config.persona} with {persona}", fg=typer.colors.YELLOW)
+            typer.secho(f"{CLI_WARNING_PREFIX}Overriding existing persona {agent_config.persona} with {persona}", fg=typer.colors.YELLOW)
             agent_config.persona = persona
             # raise ValueError(f"Cannot override {agent_config.name} existing persona {agent_config.persona} with {persona}")
         if human and human != agent_config.human:
-            typer.secho(f"Warning: Overriding existing human {agent_config.human} with {human}", fg=typer.colors.YELLOW)
+            typer.secho(f"{CLI_WARNING_PREFIX}Overriding existing human {agent_config.human} with {human}", fg=typer.colors.YELLOW)
             agent_config.human = human
             # raise ValueError(f"Cannot override {agent_config.name} existing human {agent_config.human} with {human}")
 
         # Allow overriding model specifics (model, model wrapper, model endpoint IP + type, context_window)
         if model and model != agent_config.model:
-            typer.secho(f"Warning: Overriding existing model {agent_config.model} with {model}", fg=typer.colors.YELLOW)
+            typer.secho(f"{CLI_WARNING_PREFIX}Overriding existing model {agent_config.model} with {model}", fg=typer.colors.YELLOW)
             agent_config.model = model
         if context_window is not None and int(context_window) != agent_config.context_window:
             typer.secho(
-                f"Warning: Overriding existing context window {agent_config.context_window} with {context_window}", fg=typer.colors.YELLOW
+                f"{CLI_WARNING_PREFIX}Overriding existing context window {agent_config.context_window} with {context_window}",
+                fg=typer.colors.YELLOW,
             )
             agent_config.context_window = context_window
         if model_wrapper and model_wrapper != agent_config.model_wrapper:
             typer.secho(
-                f"Warning: Overriding existing model wrapper {agent_config.model_wrapper} with {model_wrapper}", fg=typer.colors.YELLOW
+                f"{CLI_WARNING_PREFIX}Overriding existing model wrapper {agent_config.model_wrapper} with {model_wrapper}",
+                fg=typer.colors.YELLOW,
             )
             agent_config.model_wrapper = model_wrapper
         if model_endpoint and model_endpoint != agent_config.model_endpoint:
             typer.secho(
-                f"Warning: Overriding existing model endpoint {agent_config.model_endpoint} with {model_endpoint}", fg=typer.colors.YELLOW
+                f"{CLI_WARNING_PREFIX}Overriding existing model endpoint {agent_config.model_endpoint} with {model_endpoint}",
+                fg=typer.colors.YELLOW,
             )
             agent_config.model_endpoint = model_endpoint
         if model_endpoint_type and model_endpoint_type != agent_config.model_endpoint_type:
             typer.secho(
-                f"Warning: Overriding existing model endpoint type {agent_config.model_endpoint_type} with {model_endpoint_type}",
+                f"{CLI_WARNING_PREFIX}Overriding existing model endpoint type {agent_config.model_endpoint_type} with {model_endpoint_type}",
                 fg=typer.colors.YELLOW,
             )
             agent_config.model_endpoint_type = model_endpoint_type

memgpt/constants.py

@@ -17,6 +17,8 @@
 ]
 INITIAL_BOOT_MESSAGE_SEND_MESSAGE_FIRST_MSG = STARTUP_QUOTES[2]
 
+CLI_WARNING_PREFIX = "Warning: "
+
 # Constants to do with summarization / conversation length window
 # The max amount of tokens supported by the underlying model (eg 8k for gpt-4 and Mistral 7B)
 LLM_MAX_TOKENS = {

memgpt/functions/functions.py

@@ -5,7 +5,7 @@
 
 
 from memgpt.functions.schema_generator import generate_schema
-from memgpt.constants import MEMGPT_DIR
+from memgpt.constants import MEMGPT_DIR, CLI_WARNING_PREFIX
 
 sys.path.append(os.path.join(MEMGPT_DIR, "functions"))
 

memgpt/interface.py

@@ -5,6 +5,7 @@
 from colorama import Fore, Style, init
 
 from memgpt.utils import printd
+from memgpt.constants import CLI_WARNING_PREFIX
 
 init(autoreset=True)
 
@@ -108,7 +109,7 @@ def printd_user_message(icon, msg):
                 try:
                     msg_json = json.loads(msg)
                 except:
-                    printd(f"Warning: failed to parse user message into json")
+                    printd(f"{CLI_WARNING_PREFIX}failed to parse user message into json")
                     printd_user_message("🧑", msg)
                     return
         if msg_json["type"] == "user_message":
@@ -186,7 +187,7 @@ def printd_function_message(icon, msg, color=Fore.RED):
                             printd(msg_dict)
                             pass
                 else:
-                    printd(f"Warning: did not recognize function message")
+                    printd(f"{CLI_WARNING_PREFIX}did not recognize function message")
                     printd_function_message("", msg)
         else:
             try:
@@ -196,7 +197,7 @@ def printd_function_message(icon, msg, color=Fore.RED):
                 else:
                     printd_function_message("", str(msg), color=Fore.RED)
             except Exception:
-                print(f"Warning: did not recognize function message {type(msg)} {msg}")
+                print(f"{CLI_WARNING_PREFIX}did not recognize function message {type(msg)} {msg}")
                 printd_function_message("", msg)
 
     @staticmethod

memgpt/local_llm/chat_completion_proxy.py

@@ -18,6 +18,7 @@
 from memgpt.local_llm.utils import get_available_wrappers
 from memgpt.prompts.gpt_summarize import SYSTEM as SUMMARIZE_SYSTEM_MESSAGE
 from memgpt.errors import LocalLLMConnectionError, LocalLLMError
+from memgpt.constants import CLI_WARNING_PREFIX
 
 DEBUG = False
 # DEBUG = True
@@ -54,7 +55,7 @@ def get_chat_completion(
         # Warn the user that we're using the fallback
         if not has_shown_warning:
             print(
-                f"Warning: no wrapper specified for local LLM, using the default wrapper (you can remove this warning by specifying the wrapper with --wrapper)"
+                f"{CLI_WARNING_PREFIX}no wrapper specified for local LLM, using the default wrapper (you can remove this warning by specifying the wrapper with --wrapper)"
             )
             has_shown_warning = True
         if endpoint_type in ["koboldcpp", "llamacpp", "webui"]:
@@ -72,7 +73,7 @@ def get_chat_completion(
             grammar_name = "json_func_calls_with_inner_thoughts"
 
     if grammar_name is not None and endpoint_type not in ["koboldcpp", "llamacpp", "webui"]:
-        print(f"Warning: grammars are currently only supported when using llama.cpp as the MemGPT local LLM backend")
+        print(f"{CLI_WARNING_PREFIX}grammars are currently only supported when using llama.cpp as the MemGPT local LLM backend")
 
     # First step: turn the message sequence into a prompt that the model expects
     try:

memgpt/openai_tools.py

@@ -9,6 +9,7 @@
 from box import Box
 
 from memgpt.local_llm.chat_completion_proxy import get_chat_completion
+from memgpt.constants import CLI_WARNING_PREFIX
 
 
 def is_context_overflow_error(exception):
@@ -294,7 +295,10 @@ def wrapper(*args, **kwargs):
                     delay *= exponential_base * (1 + jitter * random.random())
 
                     # Sleep for the delay
-                    printd(f"Got a rate limit error ('{http_err}') on LLM backend request, waiting {int(delay)}s then retrying...")
+                    # printd(f"Got a rate limit error ('{http_err}') on LLM backend request, waiting {int(delay)}s then retrying...")
+                    print(
+                        f"{CLI_WARNING_PREFIX}Got a rate limit error ('{http_err}') on LLM backend request, waiting {int(delay)}s then retrying..."
+                    )
                     time.sleep(delay)
                 else:
                     # For other HTTP errors, re-raise the exception