fix: debug logs in server (#1452) (#1457)

Co-authored-by: Ethan Knox <ethan.m.knox@gmail.com>

.gitignore

@@ -1016,3 +1016,4 @@ pgdata/
 ## pytest mirrors
 memgpt/.pytest_cache/
 memgpy/pytest.ini
+**/**/pytest_cache

Dockerfile

@@ -1,5 +1,5 @@
 # The builder image, used to build the virtual environment
-FROM python:3.12-bookworm as builder
+FROM python:3.12.2-bookworm as builder
 ARG MEMGPT_ENVIRONMENT=PRODUCTION
 ENV MEMGPT_ENVIRONMENT=${MEMGPT_ENVIRONMENT}
 RUN pip install poetry==1.8.2
@@ -16,13 +16,13 @@ RUN poetry lock --no-update
 RUN if [ "$MEMGPT_ENVIRONMENT" = "DEVELOPMENT"  ] ; then \
     poetry install --no-root -E "postgres server dev autogen" ; \
     else \
-    poetry install --without dev --without local --no-root -E "postgres server" && \
+    poetry install --no-root -E "postgres server" && \
     rm -rf $POETRY_CACHE_DIR ;  \
     fi
 
 
 # The runtime image, used to just run the code provided its virtual environment
-FROM python:3.12-slim-bookworm as runtime
+FROM python:3.12.2-slim-bookworm as runtime
 ARG MEMGPT_ENVIRONMENT=PRODUCTION
 ENV MEMGPT_ENVIRONMENT=${MEMGPT_ENVIRONMENT}
 ENV VIRTUAL_ENV=/app/.venv \

dev-compose.yaml

@@ -22,6 +22,7 @@ services:
     build:
       context: .
       dockerfile: Dockerfile
+      target: runtime
     depends_on:
       - memgpt_db
     ports:

development.compose.yml

@@ -14,6 +14,8 @@ services:
       - .env
     environment:
       - MEMGPT_SERVER_PASS=test_server_token
+      - WATCHFILES_FORCE_POLLING=true
+
     volumes:
       - ./memgpt:/memgpt
       - ~/.memgpt/credentials:/root/.memgpt/credentials

memgpt/cli/cli.py

@@ -19,7 +19,7 @@ from memgpt.config import MemGPTConfig
 from memgpt.constants import CLI_WARNING_PREFIX, MEMGPT_DIR
 from memgpt.credentials import MemGPTCredentials
 from memgpt.data_types import EmbeddingConfig, LLMConfig, User
-from memgpt.log import logger
+from memgpt.log import get_logger
 from memgpt.metadata import MetadataStore
 from memgpt.migrate import migrate_all_agents, migrate_all_sources
 from memgpt.server.constants import WS_DEFAULT_PORT
@@ -30,6 +30,8 @@ from memgpt.streaming_interface import (
 )
 from memgpt.utils import open_folder_in_explorer, printd
 
+logger = get_logger(__name__)
+
 
 def migrate(
     debug: Annotated[bool, typer.Option(help="Print extra tracebacks for failed migrations")] = False,

memgpt/config.py

@@ -9,7 +9,9 @@ import memgpt
 import memgpt.utils as utils
 from memgpt.constants import DEFAULT_HUMAN, DEFAULT_PERSONA, DEFAULT_PRESET, MEMGPT_DIR
 from memgpt.data_types import AgentState, EmbeddingConfig, LLMConfig
-from memgpt.log import logger
+from memgpt.log import get_logger
+
+logger = get_logger(__name__)
 
 
 # helper functions for writing to configs

memgpt/constants.py

@@ -21,17 +21,6 @@ DEFAULT_PERSONA = "sam_pov"
 DEFAULT_HUMAN = "basic"
 DEFAULT_PRESET = "memgpt_chat"
 
-# Used to isolate MemGPT logger instance from Dependant Libraries logging
-LOGGER_NAME = "MemGPT"
-LOGGER_DEFAULT_LEVEL = CRITICAL
-# Where to store the logs
-LOGGER_DIR = os.path.join(MEMGPT_DIR, "logs")
-# filename of the log
-LOGGER_FILENAME = "MemGPT.log"
-# Number of log files to rotate
-LOGGER_FILE_BACKUP_COUNT = 3
-# Max Log file size in bytes
-LOGGER_MAX_FILE_SIZE = 10485760
 # LOGGER_LOG_LEVEL is use to convert Text to Logging level value for logging mostly for Cli input to setting level
 LOGGER_LOG_LEVELS = {"CRITICAL": CRITICAL, "ERROR": ERROR, "WARN": WARN, "WARNING": WARNING, "INFO": INFO, "DEBUG": DEBUG, "NOTSET": NOTSET}
 

memgpt/log.py

@@ -1,43 +1,77 @@
 import logging
-import os
-import os.path
+from logging.config import dictConfig
 from logging.handlers import RotatingFileHandler
+from pathlib import Path
+from sys import stdout
+from typing import Optional
 
-from memgpt.constants import (
-    LOGGER_DEFAULT_LEVEL,
-    LOGGER_DIR,
-    LOGGER_FILE_BACKUP_COUNT,
-    LOGGER_FILENAME,
-    LOGGER_MAX_FILE_SIZE,
-    LOGGER_NAME,
-)
+from memgpt.settings import settings
 
-# Checking if log directory exists
-if not os.path.exists(LOGGER_DIR):
-    os.makedirs(LOGGER_DIR, exist_ok=True)
+selected_log_level = logging.DEBUG if settings.debug else logging.INFO
 
-# Create logger for MemGPT
-logger = logging.getLogger(LOGGER_NAME)
-logger.setLevel(LOGGER_DEFAULT_LEVEL)
 
-# create console handler and set level to debug
-console_handler = logging.StreamHandler()
+def _setup_logfile() -> "Path":
+    """ensure the logger filepath is in place
 
-# create rotatating file handler
-file_handler = RotatingFileHandler(
-    os.path.join(LOGGER_DIR, LOGGER_FILENAME), maxBytes=LOGGER_MAX_FILE_SIZE, backupCount=LOGGER_FILE_BACKUP_COUNT
-)
+    Returns: the logfile Path
+    """
+    logfile = Path(settings.memgpt_dir / "logs" / "MemGPT.log")
+    logfile.parent.mkdir(parents=True, exist_ok=True)
+    logfile.touch(exist_ok=True)
+    return logfile
 
-# create formatters
-console_formatter = logging.Formatter("%(name)s - %(levelname)s - %(message)s")  # not datetime
-file_formatter = logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")
 
-# add formatter to console handler
-console_handler.setFormatter(console_formatter)
+# TODO: production logging should be much less invasive
+DEVELOPMENT_LOGGING = {
+    "version": 1,
+    "disable_existing_loggers": True,
+    "formatters": {
+        "standard": {"format": "%(asctime)s - %(name)s - %(levelname)s - %(message)s"},
+        "no_datetime": {
+            "format": "%(name)s - %(levelname)s - %(message)s",
+        },
+    },
+    "handlers": {
+        "console": {
+            "level": selected_log_level,
+            "class": "logging.StreamHandler",
+            "stream": stdout,
+            "formatter": "no_datetime",
+        },
+        "file": {
+            "level": "DEBUG",
+            "class": "logging.handlers.RotatingFileHandler",
+            "filename": _setup_logfile(),
+            "maxBytes": 1024**2 * 10,
+            "backupCount": 3,
+            "formatter": "standard",
+        },
+    },
+    "loggers": {
+        "MemGPT": {
+            "level": logging.DEBUG if settings.debug else logging.INFO,
+            "handlers": [
+                "console",
+                "file",
+            ],
+            "propagate": False,
+        },
+        "uvicorn": {
+            "level": "INFO",
+            "handlers": ["console"],
+            "propagate": False,
+        },
+    },
+}
 
-# add formatter for file handler
-file_handler.setFormatter(file_formatter)
 
-# add ch to logger
-logger.addHandler(console_handler)
-logger.addHandler(file_handler)
+def get_logger(name: Optional[str] = None) -> "logging.Logger":
+    """returns the project logger, scoped to a child name if provided
+    Args:
+        name: will define a child logger
+    """
+    dictConfig(DEVELOPMENT_LOGGING)
+    parent_logger = logging.getLogger("MemGPT")
+    if name:
+        return parent_logger.getChild(name)
+    return parent_logger

memgpt/metadata.py

@@ -418,7 +418,7 @@ class MetadataStore:
         """Get the user associated with a given API key"""
         token = self.get_api_key(api_key=api_key)
         if token is None:
-            raise ValueError(f"Token {api_key} does not exist")
+            raise ValueError(f"Provided token does not exist")
         else:
             return self.get_user(user_id=token.user_id)
 

memgpt/server/rest_api/auth/index.py

@@ -3,9 +3,11 @@ from uuid import UUID
 from fastapi import APIRouter, HTTPException
 from pydantic import BaseModel, Field
 
+from memgpt.log import get_logger
 from memgpt.server.rest_api.interface import QueuingInterface
 from memgpt.server.server import SyncServer
 
+logger = get_logger(__name__)
 router = APIRouter()
 
 
@@ -18,6 +20,7 @@ class AuthRequest(BaseModel):
 
 
 def setup_auth_router(server: SyncServer, interface: QueuingInterface, password: str) -> APIRouter:
+
     @router.post("/auth", tags=["auth"], response_model=AuthResponse)
     def authenticate_user(request: AuthRequest) -> AuthResponse:
         """
@@ -26,16 +29,10 @@ def setup_auth_router(server: SyncServer, interface: QueuingInterface, password:
         Currently, this is a placeholder that simply returns a UUID placeholder
         """
         interface.clear()
-        try:
-            if request.password != password:
-                # raise HTTPException(status_code=400, detail="Incorrect credentials")
-                response = server.api_key_to_user(api_key=request.password)
-            else:
-                response = server.authenticate_user()
-        except HTTPException:
-            raise
-        except Exception as e:
-            raise HTTPException(status_code=500, detail=f"{e}")
+        if request.password != password:
+            response = server.api_key_to_user(api_key=request.password)
+        else:
+            response = server.authenticate_user()
         return AuthResponse(uuid=response)
 
     return router

memgpt/server/server.py

@@ -1,5 +1,4 @@
 import json
-import logging
 import uuid
 import warnings
 from abc import abstractmethod
@@ -37,6 +36,7 @@ from memgpt.data_types import (
 # TODO use custom interface
 from memgpt.interface import AgentInterface  # abstract
 from memgpt.interface import CLIInterface  # for printing to terminal
+from memgpt.log import get_logger
 from memgpt.metadata import MetadataStore
 from memgpt.models.pydantic_models import (
     DocumentModel,
@@ -47,7 +47,7 @@ from memgpt.models.pydantic_models import (
     ToolModel,
 )
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 
 class Server(object):

memgpt/server/startup.sh

@@ -1,6 +1,7 @@
 #!/bin/sh
 echo "Starting MEMGPT server..."
 if [ "$MEMGPT_ENVIRONMENT" = "DEVELOPMENT" ] ; then
+    echo "Starting in development mode!"
     uvicorn memgpt.server.rest_api.server:app --reload --reload-dir /memgpt --host 0.0.0.0 --port 8083
 else
     uvicorn memgpt.server.rest_api.server:app --host 0.0.0.0 --port 8083

memgpt/settings.py

@@ -1,11 +1,15 @@
+from pathlib import Path
 from typing import Optional
 
+from pydantic import Field
 from pydantic_settings import BaseSettings, SettingsConfigDict
 
 
 class Settings(BaseSettings):
     model_config = SettingsConfigDict(env_prefix="memgpt_")
 
+    memgpt_dir: Optional[Path] = Field(Path.home() / ".memgpt", env="MEMGPT_DIR")
+    debug: Optional[bool] = False
     server_pass: Optional[str] = None
     pg_db: Optional[str] = None
     pg_user: Optional[str] = None

tests/test_client.py

@@ -29,7 +29,6 @@ test_server_token = "test_server_token"
 
 
 def _reset_config():
-
     # Use os.getenv with a fallback to os.environ.get
     db_url = settings.memgpt_pg_uri
 
@@ -51,14 +50,12 @@ def _reset_config():
     config.archival_storage_type = "postgres"
     config.recall_storage_type = "postgres"
     config.metadata_storage_type = "postgres"
-
     config.save()
     credentials.save()
     print("_reset_config :: ", config.config_path)
 
 
 def run_server():
-
     load_dotenv()
 
     _reset_config()

tests/test_log.py

@@ -1,16 +0,0 @@
-import logging
-
-from memgpt.constants import LOGGER_LOG_LEVELS
-from memgpt.log import logger
-
-
-def test_log_debug():
-    # test setting logging level
-    assert logging.DEBUG == LOGGER_LOG_LEVELS["DEBUG"]
-    logger.setLevel(LOGGER_LOG_LEVELS["DEBUG"])
-    assert logger.isEnabledFor(logging.DEBUG)
-
-    # Assert that the message was logged
-    assert logger.hasHandlers()
-    logger.debug("This is a Debug message")
-    assert 1 == 1