feat(proxy)!: add suppress_startup_output to turn off startup prints (#47708)

* fix(proxy): keep every startup line JSON when json_logs is on

Boot-time messages the proxy printed straight to stdout (banner, feedback
box, model list, cache and callback notices, policy table, migration
notices, CLI launch notes) and the first records of uvicorn- or
granian-spawned workers came out as plain text while json_logs was on,
breaking log pipelines that parse every line as JSON.

- print_startup_message() prints unchanged in plain mode and, with
  json_logs on, emits one INFO record on the proxy logger (ANSI
  stripped) that skips the logger and handler level checks like the
  print it replaces; show_banner() routes to it in JSON mode only
- the CLI exports JSON_LOGS=true when the config sets json_logs so
  spawned workers log JSON from their first record
- the --enforce_prisma_migration_check deprecation notice prints after
  the config is read so a config-only json_logs covers it

Resolves LIT-1624

* fix(proxy): keep show_banner() SDK-free, pick the JSON banner in initialize()

Since #46803 the `lite` shell calls show_banner() against an uninitialized
litellm package, and reading litellm.json_logs inside it initialized the whole
SDK at the banner (0.14 s to 0.83 s for a bare `lite` here, the seconds that PR
removed on small VMs). show_banner() is back to its main version; the proxy's
initialize() now emits the JSON banner record itself and calls show_banner()
only in plain mode. The banner unit tests move to the initialize() call site.

* fix(proxy): keep the JSON banner when an operator --log_config names the proxy logger

uvicorn applies --log_config after proxy_cli has switched JSON logging on and before
initialize() switches it on again through load_config(). A logging config that names
"LiteLLM Proxy" removes that logger's handlers in the window between the two, which is where
initialize() emits the banner and the feedback box, so at one worker both went nowhere;
two-worker children import litellm after uvicorn's logging setup and were not affected.
print_startup_message() now falls back to a one-off JSON handler, on the stream the JSON
handler uses, when the proxy logger has no handler of its own.

Live, json_logs on plus a log_config naming the proxy logger, one worker: main prints the
banner and the feedback box as plaintext, e2303a3925 prints neither, this head prints both as
one JSON record each. New unit test fails on e2303a3925; new integration cell
test_json_logs_with_operator_log_config_keeps_every_line_json passes at one and two workers.

* fix(proxy): keep a long startup listing whole in its JSON record

print_startup_message ran its record through the proxy logger's filters, so
in JSON mode the required-migration SQL list and the policy list were
middle-truncated at MAX_STRING_LENGTH_STDOUT_LOG (4096 by default) with the
litellm_truncated marker, where the prints they replace showed them whole.
That cap bounds echoed request payloads; a startup message is operator text,
so the record now gets the uncapped processing of a DEBUG line: secret
redaction and base64 collapse still apply, the text cap does not.

Red on 41f92b745a, green here:
- tests/unit/test_logging.py: a listing longer than the cap comes out whole;
  a credential in a startup message is still redacted
- tests/integration/configuration/test_json_logs_startup.py: a 60-policy
  config boots with json_logs on and the policy listing record (9762 chars)
  carries every policy with no truncation marker (4096 chars, marker, on
  the unfixed tree)

* fix(proxy): keep the --enforce_prisma_migration_check notice on --health and --test

This PR moved the deprecation notice after the config is read, so that a
json_logs set only in the config applies to it; --health and --test return
before that point and stopped printing it, a visible change main did not
have. The notice is now also emitted before those two returns, through the
same helper, so a deployment running `litellm --health
--enforce_prisma_migration_check` still learns the flag is dead, as on main.
A normal server start emits it once, after the config, as in the previous
commit.

Red on 2d7a2d6844, green here: tests/unit/proxy/test_proxy_cli.py asserts
the notice on --health and on --test.

* fix(proxy): keep the JSON startup records on stdout under an operator handler on the proxy logger

An operator --log_config that gives "LiteLLM Proxy" a handler of its own, a NullHandler
that silences it or a plain WARNING-level handler on stderr, is in force in the window
where the banner, the feedback box and the cache notice are emitted: uvicorn applies it
after the CLI process switched JSON on and before initialize() re-installs the JSON
handler through load_config(). 41f92b745a sent the startup records through the proxy
logger's handlers and fell back to a JSON stdout handler only when there were none, so
under such a config the records landed in the operator's handler, as plaintext on stderr,
or vanished, where the prints they replace reached stdout whatever the logging config
said.

print_startup_message() now writes every startup record through the proxy's own JSON
handler on stdout, filters included, and never through the proxy logger's handlers; the
operator's handler gets nothing from a startup message, as it got nothing from the print.
The normal JSON handler is the same class, formatter and filters, so the records a
pipeline reads are unchanged when no operator config is in play.

Tests: tests/unit/test_logging.py gets the two handler shapes (red on f6f5957d16: the
record went to stderr as plaintext or nowhere); tests/integration/configuration/
test_json_logs_startup.py gets a silencing operator --log_config cell at 1 and 2 workers
(red on f6f5957d16: banner, feedback box and cache notice missing; green here).

* fix(proxy): print the --enforce_prisma_migration_check notice when the config fails to load

f6f5957d16 printed the deprecation notice after the config is read, so a json_logs set
only in the config makes it a JSON record. A boot whose config file is missing or does
not parse then exited before that point and never printed it, where main printed it
before the error (10-20 gauntlet finding).

The notice is printed where main printed it, unless a config file can still switch
json_logs on: then it is held until the config is read and printed right after the
json_logs switch, and a click close callback prints it on the way out if the boot fails
first. functools.cache keeps it to one print either way.

Tests: tests/unit/proxy/test_proxy_cli.py covers a failed config read (red on
75043aed83), a plain config printing it once and a config-only json_logs making it one
JSON record; tests/integration/configuration/test_json_logs_startup.py adds the JSON cell
and a missing and an unparseable config cell (red on 75043aed83, green here), through
refused_boot_log, which now takes extra CLI arguments.

* feat(proxy): add suppress_startup_output to turn off startup prints

* fix(proxy): print startup messages once and without color codes off a terminal

* test(policy-engine): check the policy table keeps colors on a terminal and drops them on a pipe

* fix(proxy): keep startup warnings when suppress_startup_output is on and apply it from CONFIG_FILE_PATH and before other settings

Warnings that name something the operator must fix now print through print_startup_warning, which
skips the suppress gate. The config key is read with the CLI's json_logs handling for --config and
CONFIG_FILE_PATH, applied before the rest of litellm_settings in load_config, and can only turn the
setting on. The Prometheus config panel obeys the setting, and startup prints use the plain log
format's color rule.

* fix(proxy): read CONFIG_FILE_PATH in the CLI only when workers would load it

Workers load CONFIG_FILE_PATH only when it names an existing YAML file, so the CLI's early read of
json_logs and suppress_startup_output skips any other value instead of failing a boot that used to start.

* fix(proxy): read only suppress_startup_output from CONFIG_FILE_PATH in the CLI, without secrets

The CLI now reads CONFIG_FILE_PATH as a plain file with its includes, skips it when a config bucket is set,
and applies only suppress_startup_output, also when --config is passed. json_logs there stays with the
workers as before. Fatal boot errors drop their color codes off a terminal like the startup prints.

* fix(logging): emit startup warnings as WARNING records in JSON mode

* fix(proxy): keep one JSON record per search tool in the startup listing

* fix(proxy): keep booting with an ASCII stdout or an empty litellm_settings

Startup messages now go through click.echo, which writes UTF-8 when stdout is ASCII, as the banner did
before. An empty litellm_settings section reads as no settings in the CLI, as the workers already read it

* test(proxy): boot the CLI tests on the real ProxyConfig config readers

* test(proxy): cover the startup warnings that print with startup output suppressed

* test(proxy): boot the deprecated resolver flag test on the real ProxyConfig

* test(proxy): drop the startup output contract's narrating docstrings

* test(proxy): cover the pgbouncer start failure message off a terminal

---------

Co-authored-by: yucheng-berri <293926326+yucheng-berri@users.noreply.github.com>

litellm/__init__.py

@@ -39,6 +39,7 @@ from litellm._logging import (
     turn_on_json,
     log_level,
 )
+from litellm._logging import suppress_startup_output as suppress_startup_output
 import re
 from litellm.constants import (
     DEFAULT_BATCH_SIZE,

litellm/_logging.py

@@ -10,10 +10,12 @@ import sys
 from collections.abc import Iterator, Mapping
 from datetime import datetime
 from logging import Formatter
+from types import FrameType
 from typing import Final, TextIO
 from urllib.parse import unquote
 
 from pydantic import TypeAdapter
+from typing_extensions import ReadOnly, TypedDict
 
 import litellm
 from litellm.constants import (
@@ -598,6 +600,7 @@ _LOG_FORMAT_PREFIX: Final = "%(asctime)s - %(name)s:%(levelname)s"
 _LOG_FORMAT_SUFFIX: Final = ": %(filename)s:%(lineno)s - %(message)s"
 _PLAIN_LOG_FORMAT: Final = _LOG_FORMAT_PREFIX + _LOG_FORMAT_SUFFIX
 _COLOR_LOG_FORMAT: Final = f"\033[92m{_LOG_FORMAT_PREFIX}\033[0m{_LOG_FORMAT_SUFFIX}"
+_ANSI_ESCAPE: Final = re.compile(r"\x1b\[[0-9;]*m")
 
 
 def _stream_is_tty(stream: TextIO | None) -> bool:
@@ -612,15 +615,28 @@ def _stream_is_tty(stream: TextIO | None) -> bool:
         return False
 
 
-def _plain_log_format(stdout: TextIO | None, stderr: TextIO | None) -> str:
-    """The plain-text log format, colorized only when both streams are an interactive terminal.
+def _color_enabled(stream: TextIO | None) -> bool:
+    """True when ANSI colors written to the stream reach an interactive terminal.
 
     Honors the NO_COLOR convention from no-color.org: color is disabled when
     NO_COLOR is present with a non-empty value.
     """
-    if os.environ.get("NO_COLOR"):
-        return _PLAIN_LOG_FORMAT
-    return _COLOR_LOG_FORMAT if _stream_is_tty(stdout) and _stream_is_tty(stderr) else _PLAIN_LOG_FORMAT
+    return not os.environ.get("NO_COLOR") and _stream_is_tty(stream)
+
+
+def _terminal_colors(stdout: TextIO | None, stderr: TextIO | None) -> bool:
+    """True when both streams are an interactive color terminal, the one rule for log lines and startup prints."""
+    return _color_enabled(stdout) and _color_enabled(stderr)
+
+
+def terminal_text(message: str) -> str:
+    """The message with its ANSI colors kept only on a color terminal, by the rule the plain log format uses."""
+    return message if _terminal_colors(sys.stdout, sys.stderr) else _ANSI_ESCAPE.sub("", message)
+
+
+def _plain_log_format(stdout: TextIO | None, stderr: TextIO | None) -> str:
+    """The plain-text log format, colorized only when both streams are an interactive terminal."""
+    return _COLOR_LOG_FORMAT if _terminal_colors(stdout, stderr) else _PLAIN_LOG_FORMAT
 
 
 class LevelRoutingStreamHandler(logging.StreamHandler):
@@ -642,8 +658,8 @@ class LevelRoutingStreamHandler(logging.StreamHandler):
         super().emit(record)
 
 
-def _parse_json_logs_env(value: str | None) -> bool:
-    """Strict opt-in parse for the JSON_LOGS env var: only "true" (any case) enables JSON logs.
+def _parse_env_flag(value: str | None) -> bool:
+    """Strict opt-in parse for a boolean env var such as JSON_LOGS: only "true" (any case) enables it.
 
     Matches the reader in litellm-proxy-extras/_logging.py. The previous
     bool(os.getenv(...)) treated any non-empty value, including "false" and "0",
@@ -656,7 +672,18 @@ def resolve_log_level(log_level: str) -> int:
     return getattr(logging, log_level.upper())
 
 
-json_logs: Final = _parse_json_logs_env(os.getenv("JSON_LOGS"))
+json_logs: Final = _parse_env_flag(os.getenv("JSON_LOGS"))
+suppress_startup_output: Final = _parse_env_flag(os.getenv("LITELLM_SUPPRESS_STARTUP_OUTPUT"))
+
+
+class StartupOutputSettings(TypedDict, total=False):
+    json_logs: ReadOnly[object]
+    suppress_startup_output: ReadOnly[object]
+
+
+STARTUP_OUTPUT_SETTINGS: Final = TypeAdapter(StartupOutputSettings)
+
+
 # Create a handler for the logger (you may need to adapt this based on your needs)
 log_level: Final = os.getenv("LITELLM_LOG", "DEBUG")
 numeric_level: Final[int] = resolve_log_level(log_level)
@@ -743,7 +770,7 @@ class JsonFormatter(Formatter):
         return dt.isoformat()
 
     def format(self, record):
-        message_str: Final = record.getMessage()
+        message_str: Final = _ANSI_ESCAPE.sub("", record.getMessage())
         json_record: Final[dict[str, object]] = {
             "message": message_str,
             "level": record.levelname,
@@ -799,12 +826,14 @@ class CorrelationPlainFormatter(logging.Formatter):
     """Appends trace_id/session_id to plain-text log lines stamped by CorrelationContextFilter.
 
     Mirrors JsonFormatter's handling of these two fields so request_correlation_in_logs
-    behaves the same whether or not json_logs is enabled.
+    behaves the same whether or not json_logs is enabled. Unless it was built with the color format,
+    it also drops the ANSI color codes a message carries, which a piped log would store as text.
     """
 
     def format(self, record: logging.LogRecord) -> str:
         rendered: Final = super().format(record)
-        formatted: Final = rendered if _is_redacted(record) else redact_string(rendered)
+        uncolored: Final = rendered if self._fmt == _COLOR_LOG_FORMAT else _ANSI_ESCAPE.sub("", rendered)
+        formatted: Final = uncolored if _is_redacted(record) else redact_string(uncolored)
         trace_id: Final = getattr(record, "trace_id", None)
         session_id: Final = getattr(record, "session_id", None)
         if not trace_id and not session_id:
@@ -898,29 +927,46 @@ verbose_proxy_logger.addFilter(_diagnostic_filter)
 verbose_proxy_stdout_logger.addFilter(_diagnostic_filter)
 verbose_logger.addFilter(_diagnostic_filter)
 
-_ANSI_ESCAPE: Final = re.compile(r"\x1b\[[0-9;]*m")
-
 
 def print_startup_message(message: str) -> None:
     """Write a boot-time message meant for the operator's terminal.
 
-    Plain mode prints it unchanged, colors and spacing included. With JSON logs on, every line of
-    stdout is read by a log pipeline, so the message goes out as one INFO record of the proxy logger
-    instead, on stdout like the print whatever handlers the logging config gave that logger, with the
-    ANSI color codes and surrounding blank lines removed, located at the caller and emitted at any log
-    level, whole and unredacted as the print showed it; a blank spacer produces no record.
+    LITELLM_SUPPRESS_STARTUP_OUTPUT or litellm_settings.suppress_startup_output turns every one of them
+    off; a warning goes through print_startup_warning instead. Plain mode prints it with its spacing,
+    keeping the ANSI colors only on a color terminal by the same check as the plain log format, so a
+    container log never stores escape codes.
+    With JSON logs on, every line of stdout is read by a log pipeline, so the message goes out as one
+    INFO record of the proxy logger instead, on stdout like the print whatever handlers the logging
+    config gave that logger, with the ANSI color codes and surrounding blank lines removed, located at
+    the caller and emitted at any log level, whole and unredacted as the print showed it; a blank
+    spacer produces no record.
     """
+    if litellm.suppress_startup_output is True:
+        return
+    frame: Final = inspect.currentframe()
+    _write_startup_message(message, frame.f_back if frame is not None else None, logging.INFO)
+
+
+def print_startup_warning(message: str) -> None:
+    """Write a boot-time warning as print_startup_message writes a message, even with suppress_startup_output
+    on, since it names something the operator has to fix. With JSON logs on it is a WARNING record, on
+    stdout where the print put it, so a collector that filters on the level field keeps it."""
+    frame: Final = inspect.currentframe()
+    _write_startup_message(message, frame.f_back if frame is not None else None, logging.WARNING)
+
+
+def _write_startup_message(message: str, caller: FrameType | None, level: int) -> None:
     if not litellm.json_logs:
-        print(message)  # noqa: T201  # plain mode: this is the operator-terminal print it replaces
+        import click
+
+        click.echo(terminal_text(message))
         return
     text: Final = _ANSI_ESCAPE.sub("", message).strip("\n")
     if not text:
         return
-    frame: Final = inspect.currentframe()
-    caller: Final = frame.f_back if frame is not None else None
     record: Final = verbose_proxy_logger.makeRecord(
         verbose_proxy_logger.name,
-        logging.INFO,
+        level,
         caller.f_code.co_filename if caller is not None else __file__,
         caller.f_lineno if caller is not None else 0,
         text,
@@ -934,8 +980,8 @@ def print_startup_message(message: str) -> None:
 
 
 def _startup_json_handler() -> logging.Handler:
-    """The proxy's JSON handler on stdout, filters included, for one startup message."""
-    handler: Final = LevelRoutingStreamHandler()
+    """The proxy's JSON handler on stdout, filters included, for one startup message or warning."""
+    handler: Final = logging.StreamHandler(sys.stdout)
     handler.setFormatter(JsonFormatter())
     handler.addFilter(_secret_filter)
     handler.addFilter(_correlation_filter)

litellm/integrations/prometheus.py

@@ -1173,8 +1173,8 @@ class PrometheusLogger(CustomLogger):
         # Build label filters from valid configurations
         label_filters: Final = self._build_label_filters(parsed_configs)
 
-        # Pretty print the processed configuration
-        self._pretty_print_prometheus_config(label_filters)
+        if litellm.suppress_startup_output is not True:
+            self._pretty_print_prometheus_config(label_filters)
         return label_filters
 
     def _parse_exclude_config(self) -> tuple[frozenset[str], frozenset[str]]:

litellm/proxy/db/check_migration.py

@@ -4,7 +4,7 @@ import os
 import subprocess
 from typing import Final
 
-from litellm._logging import print_startup_message, verbose_logger
+from litellm._logging import print_startup_message, print_startup_warning, verbose_logger
 
 
 def extract_sql_commands(diff_output: str) -> list[str]:
@@ -63,7 +63,7 @@ def check_prisma_schema_diff_helper(db_url: str) -> tuple[bool, list[str]]:
             run_prisma,
         )
     except ImportError as e:
-        print_startup_message(
+        print_startup_warning(
             f"Skipping the migration diff: litellm-proxy-extras has no Prisma runner. Error: {e}",
         )
         return False, []
@@ -96,14 +96,14 @@ def check_prisma_schema_diff_helper(db_url: str) -> tuple[bool, list[str]]:
         else:
             return False, []
     except subprocess.TimeoutExpired:
-        print_startup_message(
+        print_startup_warning(
             f"Timed out after {timeout}s generating the migration diff. "
             f"Raise {PRISMA_COMMAND_TIMEOUT_ENV_VAR} if this database needs longer.",
         )
         return False, []
     except subprocess.CalledProcessError as e:
         error_message: Final = f"Failed to generate migration diff. Error: {e.stderr}"
-        print_startup_message(error_message)
+        print_startup_warning(error_message)
         return False, []
 
 

litellm/proxy/policy_engine/init_policies.py

@@ -11,7 +11,7 @@ from collections.abc import Mapping, Sequence
 from itertools import chain
 from typing import TYPE_CHECKING, Any, Final, Optional
 
-from litellm._logging import print_startup_message, verbose_proxy_logger
+from litellm._logging import print_startup_message, print_startup_warning, verbose_proxy_logger
 from litellm.proxy.policy_engine.attachment_registry import get_attachment_registry
 from litellm.proxy.policy_engine.policy_registry import get_policy_registry
 from litellm.proxy.policy_engine.policy_validator import PolicyValidator
@@ -93,7 +93,7 @@ def _print_policies_on_startup(
         )
     else:
         print_startup_message("\n".join((header, *policy_lines)))
-        print_startup_message(
+        print_startup_warning(
             f"\n{_yellow_color_code}Warning: No policy_attachments configured. Policies will not be applied to any requests.{_reset_color_code}\n",
         )
     sys.stdout.flush()

litellm/proxy/proxy_cli.py

@@ -10,16 +10,17 @@ import sys
 import urllib.parse as urlparse
 from collections.abc import Callable, Iterable, Mapping, Sequence
 from pathlib import Path
-from typing import TYPE_CHECKING, Any, Final
+from typing import TYPE_CHECKING, Annotated, Any, Final
 
 import click
 import httpx
 from click.core import ParameterSource
 from dotenv import load_dotenv
-from pydantic import ConfigDict
+from pydantic import BeforeValidator, ConfigDict, TypeAdapter
+from typing_extensions import ReadOnly, TypedDict
 
 import litellm
-from litellm._logging import print_startup_message
+from litellm._logging import StartupOutputSettings, print_startup_message, print_startup_warning, terminal_text
 from litellm.constants import DEFAULT_NUM_WORKERS_LITELLM_PROXY
 from litellm.proxy.db.pgbouncer import (
     PgBouncerError,
@@ -205,7 +206,7 @@ def deprecated_v2_flag_passed_on_cli() -> bool:
 
 
 def _print_enforce_prisma_migration_check_notice() -> None:
-    print_startup_message(
+    print_startup_warning(
         "\033[1;33mLiteLLM Proxy: --enforce_prisma_migration_check is "
         "deprecated and has no effect, because the proxy always exits "
         "when database setup fails at startup. You can safely remove it.\033[0m",
@@ -227,6 +228,36 @@ def _print_or_hold_enforce_prisma_migration_check_notice(
     return print_once
 
 
+def _empty_as_none(value: object) -> object:
+    return value or None
+
+
+class _ConfigStartupOutputSettings(TypedDict, total=False):
+    litellm_settings: ReadOnly[Annotated[StartupOutputSettings | None, BeforeValidator(_empty_as_none)]]
+
+
+_CONFIG_STARTUP_OUTPUT_SETTINGS: Final = TypeAdapter(_ConfigStartupOutputSettings)
+
+
+def _apply_config_suppress_startup_output(litellm_settings: StartupOutputSettings | None) -> None:
+    if litellm_settings is None or litellm_settings.get("suppress_startup_output") is not True:
+        return
+    litellm.suppress_startup_output = True
+    os.environ["LITELLM_SUPPRESS_STARTUP_OUTPUT"] = "true"
+
+
+def _apply_config_logging_settings(litellm_settings: StartupOutputSettings | None) -> None:
+    """Turn on the config's json_logs and suppress_startup_output before any worker starts. Workers that
+    uvicorn or granian spawn import litellm afresh and read both from the env at import, so they log JSON
+    and stay quiet from their first line instead of from when load_config reaches the key."""
+    _apply_config_suppress_startup_output(litellm_settings)
+    if litellm_settings is None or litellm_settings.get("json_logs") is not True:
+        return
+    litellm.json_logs = True
+    os.environ["JSON_LOGS"] = "true"
+    litellm.turn_on_json()
+
+
 class ProxyInitializationHelpers:
     @staticmethod
     def echo_litellm_version():
@@ -373,7 +404,7 @@ class ProxyInitializationHelpers:
             if "timeout_worker_healthcheck" in inspect.signature(uvicorn.Config.__init__).parameters:
                 uvicorn_args["timeout_worker_healthcheck"] = timeout_worker_healthcheck
             else:
-                print_startup_message(
+                print_startup_warning(
                     f"\033[1;33mLiteLLM Proxy: --timeout_worker_healthcheck "
                     f"requires uvicorn>=0.37.0, but installed uvicorn=={uvicorn.__version__}. "
                     f"Ignoring the flag.\033[0m",
@@ -396,7 +427,7 @@ class ProxyInitializationHelpers:
         import uvicorn
 
         if max_requests_before_restart is None:
-            print_startup_message(
+            print_startup_warning(
                 "\033[1;33mLiteLLM Proxy: --max_requests_before_restart_jitter "
                 "has no effect without --max_requests_before_restart\033[0m\n",
             )
@@ -404,7 +435,7 @@ class ProxyInitializationHelpers:
         if "limit_max_requests_jitter" in inspect.signature(uvicorn.Config.__init__).parameters:
             uvicorn_args["limit_max_requests_jitter"] = jitter
         else:
-            print_startup_message(
+            print_startup_warning(
                 f"\033[1;33mLiteLLM Proxy: --max_requests_before_restart_jitter "
                 f"requires uvicorn>=0.41.0, but installed uvicorn=={uvicorn.__version__}. "
                 f"Ignoring the flag.\033[0m",
@@ -571,12 +602,12 @@ class ProxyInitializationHelpers:
 
         print_startup_message(f"\033[1;32mLiteLLM Proxy: Starting server on {host}:{port} using Granian\033[0m\n")
         if max_requests_before_restart is not None:
-            print_startup_message(
+            print_startup_warning(
                 "\033[1;33mLiteLLM: --max_requests_before_restart is not supported by Granian "
                 "(Granian uses workers_lifetime in seconds, not a per-request limit).\033[0m\n",
             )
         if ciphers is not None:
-            print_startup_message("\033[1;33mLiteLLM: --ciphers is not applied when using --run_granian.\033[0m\n")
+            print_startup_warning("\033[1;33mLiteLLM: --ciphers is not applied when using --run_granian.\033[0m\n")
 
         kwargs: Final[dict[str, object]] = {
             "target": "litellm.proxy.proxy_server:app",
@@ -698,7 +729,7 @@ class ProxyInitializationHelpers:
             gunicorn_options["max_requests"] = max_requests_before_restart
         if max_requests_before_restart_jitter is not None:
             if max_requests_before_restart is None:
-                print_startup_message(
+                print_startup_warning(
                     "\033[1;33mLiteLLM Proxy: --max_requests_before_restart_jitter "
                     "has no effect without --max_requests_before_restart\033[0m\n",
                 )
@@ -739,7 +770,7 @@ class ProxyInitializationHelpers:
             with open(os.devnull, "w") as devnull:
                 subprocess.Popen(command, stdout=devnull, stderr=devnull)
         except Exception as e:
-            print_startup_message(f"""
+            print_startup_warning(f"""
                 LiteLLM Warning: proxy started with `ollama` model\n`ollama serve` failed with Exception{e}. \nEnsure you run `ollama serve`
             """)
 
@@ -1325,6 +1356,19 @@ def run_server(
                 os.environ[k] = v
 
         litellm_settings = None
+        env_config_file: Final = os.environ.get("CONFIG_FILE_PATH")
+        if (
+            env_config_file
+            and os.environ.get("LITELLM_CONFIG_BUCKET_NAME") is None
+            and ProxyConfig().is_yaml(config_file_path=env_config_file)
+        ):
+            import asyncio
+
+            _apply_config_suppress_startup_output(
+                _CONFIG_STARTUP_OUTPUT_SETTINGS.validate_python(
+                    asyncio.run(ProxyConfig().read_config_file(config_file_path=env_config_file))
+                ).get("litellm_settings")
+            )
         if config is not None:
             """
             Allow user to pass in db url via config
@@ -1342,17 +1386,9 @@ def run_server(
 
             ### LITELLM SETTINGS ###
             litellm_settings = _config.get("litellm_settings", None)
-            if (
-                litellm_settings is not None
-                and "json_logs" in litellm_settings
-                and litellm_settings["json_logs"] is True
-            ):
-                import litellm
-
-                litellm.json_logs = True
-                os.environ["JSON_LOGS"] = "true"
-
-                litellm.turn_on_json()
+            _apply_config_logging_settings(
+                _CONFIG_STARTUP_OUTPUT_SETTINGS.validate_python(_config).get("litellm_settings")
+            )
             if held_prisma_notice is not None:
                 held_prisma_notice()
             ### GENERAL SETTINGS ###
@@ -1450,7 +1486,9 @@ def run_server(
                 _bad_scheme = unsupported_db_scheme(_candidate_url)
                 if _bad_scheme is not None:
                     print(
-                        f"\033[1;31mLiteLLM Proxy: {unsupported_db_scheme_message(_db_env, _bad_scheme)}\033[0m",
+                        terminal_text(
+                            f"\033[1;31mLiteLLM Proxy: {unsupported_db_scheme_message(_db_env, _bad_scheme)}\033[0m"
+                        ),
                         file=sys.stderr,
                         flush=True,
                     )
@@ -1553,7 +1591,7 @@ def run_server(
                         env_value=os.getenv("USE_V2_MIGRATION_RESOLVER"),
                     )
                     if deprecated_v2_flag_passed_on_cli() and use_v2_resolver:
-                        print_startup_message(
+                        print_startup_warning(
                             "\033[1;33mLiteLLM Proxy: --use_v2_migration_resolver is "
                             "deprecated and has no effect, because the v2 migration "
                             "resolver is now the default. You can safely remove it. To "
@@ -1561,7 +1599,7 @@ def run_server(
                             "--use_legacy_migration_resolver.\033[0m",
                         )
                     if not use_v2_resolver:
-                        print_startup_message(
+                        print_startup_warning(
                             "\033[1;33mLiteLLM Proxy: Using the legacy (v1) migration "
                             "resolver. It performs the diff-and-force recovery that can "
                             "cause schema thrashing during rolling deploys where two "
@@ -1583,22 +1621,26 @@ def run_server(
                         # issues, and any `prisma db push` against a
                         # partitioned LiteLLM_SpendLogs.
                         print(
-                            f"\033[1;31mLiteLLM Proxy: Database migration cannot proceed. {e}\033[0m",
+                            terminal_text(f"\033[1;31mLiteLLM Proxy: Database migration cannot proceed. {e}\033[0m"),
                             file=sys.stderr,
                             flush=True,
                         )
                         sys.exit(2)
                     if not setup_ok:
                         print(
-                            "\033[1;31mLiteLLM Proxy: Database setup failed after multiple retries. "
-                            "The proxy cannot start safely. Please check your database connection and migration status.\033[0m"
+                            terminal_text(
+                                "\033[1;31mLiteLLM Proxy: Database setup failed after multiple retries. "
+                                "The proxy cannot start safely. Please check your database connection and migration status.\033[0m"
+                            )
                         )
                         sys.exit(1)
             else:
                 print(
-                    "\033[1;31mLiteLLM Proxy: a database URL is set but the prisma CLI is neither on PATH nor importable "
-                    "as a package, so the database cannot be set up. Install it with `pip install 'litellm[extra_proxy]'` "
-                    "or run a shipped LiteLLM image.\033[0m"
+                    terminal_text(
+                        "\033[1;31mLiteLLM Proxy: a database URL is set but the prisma CLI is neither on PATH nor importable "
+                        "as a package, so the database cannot be set up. Install it with `pip install 'litellm[extra_proxy]'` "
+                        "or run a shipped LiteLLM image.\033[0m"
+                    )
                 )
                 sys.exit(1)
         pgbouncer_settings: Final = PgBouncerSettings()
@@ -1609,8 +1651,10 @@ def run_server(
             )
             if isinstance(pooled_database_url, PgBouncerError):
                 print(
-                    f"\033[1;31mLiteLLM Proxy: LITELLM_PGBOUNCER_ENABLED is set but the in-container pgbouncer "
-                    f"could not start: {pooled_database_url.reason}\033[0m",
+                    terminal_text(
+                        f"\033[1;31mLiteLLM Proxy: LITELLM_PGBOUNCER_ENABLED is set but the in-container pgbouncer "
+                        f"could not start: {pooled_database_url.reason}\033[0m"
+                    ),
                     file=sys.stderr,
                     flush=True,
                 )
@@ -1675,7 +1719,7 @@ def run_server(
                 run_gunicorn=run_gunicorn is True,
             )
             if single_worker_warning is not None:
-                print_startup_message(f"\033[1;33m{single_worker_warning}\033[0m\n")
+                print_startup_warning(f"\033[1;33m{single_worker_warning}\033[0m\n")
         if run_gunicorn is False and run_hypercorn is False and run_granian is False:
             if limit_concurrency is not None:
                 uvicorn_args["limit_concurrency"] = limit_concurrency

litellm/proxy/proxy_server.py

@@ -310,6 +310,7 @@ import litellm._redis
 from litellm import Router
 from litellm._internal_context import service_target, with_service_target
 from litellm._logging import (
+    STARTUP_OUTPUT_SETTINGS,
     print_startup_message,
     redact_string,
     verbose_proxy_logger,
@@ -5805,6 +5806,12 @@ class ProxyConfig:
         # verbose_proxy_logger.debug(f"loaded config={json.dumps(config, indent=4)}")
         return config
 
+    async def read_config_file(self, config_file_path: str) -> Mapping[str, object]:
+        """The config file with its includes merged, without the bucket, database and secret steps of get_config."""
+        return await self._process_includes(
+            config=self._load_yaml_file(config_file_path), config_file_path=os.path.abspath(config_file_path)
+        )
+
     async def _process_includes(self, config: dict, config_file_path: str) -> dict:
         """
         Process includes by appending their contents to the main config
@@ -6429,13 +6436,8 @@ class ProxyConfig:
 
         search_tools_parsed: Final[list[SearchToolTypedDict]] = []
 
-        print_startup_message("\033[32mLiteLLM: Proxy initialized with Search Tools:\033[0m")
-
         for search_tool in search_tools_raw:
-            # Display loaded search tool
             search_tool_name = search_tool.get("search_tool_name", "")
-            search_provider = search_tool.get("litellm_params", {}).get("search_provider", "")
-            print_startup_message(f"\033[32m    {search_tool_name} ({search_provider})\033[0m")
 
             # Handle os.environ/ variables in litellm_params
             litellm_params = search_tool.get("litellm_params", {})
@@ -6605,13 +6607,15 @@ class ProxyConfig:
             from litellm.types.integrations.prometheus import validate_caller_identity_settings
 
             validate_caller_identity_settings(litellm_settings)
+            if STARTUP_OUTPUT_SETTINGS.validate_python(litellm_settings).get("suppress_startup_output") is True:
+                litellm.suppress_startup_output = True
 
             # ANSI escape code for blue text
             blue_color_code: Final = "\033[94m"
             reset_color_code: Final = "\033[0m"
             for key, value in litellm_settings.items():
                 if key == "cache" and value is True:
-                    print_startup_message(f"{blue_color_code}\nSetting Cache on Proxy{reset_color_code}")
+                    print_startup_message(f"\n{blue_color_code}Setting Cache on Proxy{reset_color_code}")
                     from litellm.caching.caching import Cache
 
                     cache_params = {}
@@ -6926,6 +6930,8 @@ class ProxyConfig:
                     setattr(litellm, key, value)
                 elif key == "drop_params":
                     litellm.drop_params = drop_params_flag(value, "litellm_settings.drop_params", verbose_proxy_logger)
+                elif key == "suppress_startup_output":
+                    continue
                 else:
                     verbose_proxy_logger.debug(
                         "%s setting litellm.%s=%s%s",
@@ -7281,6 +7287,13 @@ class ProxyConfig:
 
         ## SEARCH TOOLS SETTINGS
         search_tools: Final[list[SearchToolTypedDict] | None] = self.parse_search_tools(config)
+        if search_tools:
+            print_startup_message("\033[32mLiteLLM: Proxy initialized with Search Tools:\033[0m")
+            for tool in search_tools:
+                print_startup_message(
+                    f"\033[32m    {tool.get('search_tool_name', '')} "
+                    f"({tool.get('litellm_params', {}).get('search_provider', '')})\033[0m"
+                )
 
         ## SANDBOX TOOLS SETTINGS
         from litellm.sandbox.sandbox_tools import register_sandbox_tools
@@ -9683,12 +9696,9 @@ async def initialize(
         master_key, \
         user_custom_auth, \
         prisma_client
-    from litellm.proxy.common_utils.banner import LITELLM_BANNER, show_banner
+    from litellm.proxy.common_utils.banner import LITELLM_BANNER
 
-    if litellm.json_logs:
-        print_startup_message(LITELLM_BANNER)
-    else:
-        show_banner()
+    print_startup_message(f"\n{LITELLM_BANNER}\n")
     if os.getenv("LITELLM_DONT_SHOW_FEEDBACK_BOX", "").lower() != "true":
         generate_feedback_box()
     user_model = model

tests/integration/configuration/test_startup_output.py

@@ -0,0 +1,133 @@
+from __future__ import annotations
+
+import uuid
+from collections.abc import Mapping
+from pathlib import Path
+from typing import Final
+
+import pytest
+import yaml
+from pydantic import JsonValue
+
+from tests.integration._support.client import Gateway, eventually
+from tests.integration._support.process import owned_proxy_process
+
+STARTED_WORKER: Final = "Started server process"
+RELOAD_LINE: Final = "search tool(s) into router"
+RELOADS_PER_WORKER: Final = 3
+RELOAD_WAIT_SECONDS: Final = 60
+SEARCH_TOOLS_HEADER: Final = "Proxy initialized with Search Tools:"
+STARTUP_MESSAGES: Final = (
+    "██╗",
+    "Give Feedback / Get Help",
+    "Proxy initialized with Config, Set models:",
+    "Setting Cache on Proxy",
+    SEARCH_TOOLS_HEADER,
+    "Postgres connection budget",
+)
+ANSI_ESCAPE: Final = "\x1b["
+WORKER_COUNTS: Final = (1, 2)
+CACHE_SETTINGS: Final[Mapping[str, JsonValue]] = {"cache": True, "cache_params": {"type": "redis"}}
+
+
+def _lines_with(log: str, text: str) -> int:
+    return sum(text in line for line in log.splitlines())
+
+
+def _write_config(
+    directory: Path, upstream_url: str, model: str, search_tool: str, litellm_settings: Mapping[str, JsonValue]
+) -> Path:
+    config: Final = directory / f"startup_output_{uuid.uuid4().hex}.yaml"
+    config.write_text(
+        yaml.safe_dump(
+            {
+                "model_list": [
+                    {
+                        "model_name": model,
+                        "litellm_params": {
+                            "model": f"openai/{model}",
+                            "api_base": f"{upstream_url}/v1",
+                            "api_key": "integration-provider-key",
+                        },
+                    }
+                ],
+                "search_tools": [
+                    {
+                        "search_tool_name": search_tool,
+                        "litellm_params": {"search_provider": "tavily", "api_key": "integration-search-key"},
+                    }
+                ],
+                "litellm_settings": dict(litellm_settings),
+                "general_settings": {
+                    "master_key": "os.environ/LITELLM_MASTER_KEY",
+                    "database_url": "os.environ/DATABASE_URL",
+                    "proxy_config_reload_interval_seconds": 1,
+                },
+            },
+            sort_keys=False,
+        )
+    )
+    return config
+
+
+def _boot_reload_and_stop(
+    gateway: Gateway,
+    directory: Path,
+    litellm_settings: Mapping[str, JsonValue],
+    overrides: Mapping[str, str],
+    workers: int,
+) -> tuple[str, str]:
+    stamp: Final = uuid.uuid4().hex
+    model: Final = f"startup-output-{stamp}"
+    search_tool: Final = f"startup-output-search-{stamp}"
+    config: Final = _write_config(directory, gateway.upstream_url, model, search_tool, litellm_settings)
+    with owned_proxy_process(
+        gateway,
+        directory,
+        {"LITELLM_LOG": "INFO", **overrides},
+        config=config,
+        workers=workers,
+    ) as owned:
+        reply: Final = owned.gateway.chat(model, text=f"startup output {stamp}")
+        assert reply["model"] == model, reply
+        reloads: Final = eventually(
+            lambda: _lines_with(owned.log.read_text(), RELOAD_LINE),
+            lambda count: count >= RELOADS_PER_WORKER * workers,
+            seconds=RELOAD_WAIT_SECONDS,
+            return_last_on_timeout=True,
+        )
+        assert reloads >= RELOADS_PER_WORKER * workers, reloads
+        log: Final = owned.log
+    return log.read_text(), search_tool
+
+
+@pytest.mark.parametrize("workers", WORKER_COUNTS)
+def test_startup_messages_reach_a_piped_stdout_once_and_without_color_codes(
+    gateway: Gateway, tmp_path: Path, workers: int
+) -> None:
+    log, search_tool = _boot_reload_and_stop(gateway, tmp_path, CACHE_SETTINGS, {}, workers)
+
+    for message in STARTUP_MESSAGES:
+        assert message in log, (message, log)
+    assert _lines_with(log, SEARCH_TOOLS_HEADER) == workers, log
+    assert _lines_with(log, f"{search_tool} (tavily)") == workers, log
+    colored: Final = tuple(line for line in log.splitlines() if ANSI_ESCAPE in line)
+    assert colored == (), "lines with ANSI color codes on a piped stdout:\n" + "\n".join(colored)
+
+
+@pytest.mark.parametrize(
+    ("litellm_settings", "overrides"),
+    [
+        pytest.param(CACHE_SETTINGS, {"LITELLM_SUPPRESS_STARTUP_OUTPUT": "true"}, id="env-plain"),
+        pytest.param({**CACHE_SETTINGS, "json_logs": True, "suppress_startup_output": True}, {}, id="config-json"),
+    ],
+)
+def test_suppress_startup_output_leaves_out_every_startup_message(
+    gateway: Gateway, tmp_path: Path, litellm_settings: Mapping[str, JsonValue], overrides: Mapping[str, str]
+) -> None:
+    workers: Final = 2
+    log, search_tool = _boot_reload_and_stop(gateway, tmp_path, litellm_settings, overrides, workers)
+
+    assert _lines_with(log, STARTED_WORKER) == workers, log
+    printed: Final = tuple(message for message in (*STARTUP_MESSAGES, search_tool) if message in log)
+    assert printed == (), (printed, log)

tests/unit/enterprise/integrations/test_prometheus.py

@@ -217,6 +217,26 @@ def test_future_metrics_with_end_user_are_filtered():
         litellm.enable_end_user_cost_tracking_prometheus_only = original_setting
 
 
+@pytest.mark.parametrize("suppressed", [False, True])
+def test_prometheus_config_panel_obeys_suppress_startup_output(
+    monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str], suppressed: bool
+):
+    """The processed-config panel is a startup print like the banner, so suppress_startup_output turns it off
+    while the label filters still load."""
+    clear_prometheus_registry()
+    monkeypatch.setattr(litellm, "suppress_startup_output", suppressed)
+    monkeypatch.setattr(
+        litellm,
+        "prometheus_metrics_config",
+        [{"group": "service_metrics", "metrics": ["litellm_deployment_total_requests"], "include_labels": ["team"]}],
+    )
+
+    label_filters = PrometheusLogger()._parse_prometheus_config()
+
+    assert label_filters["litellm_deployment_total_requests"] == ["team"]
+    assert ("Prometheus Configuration Processed" in capfd.readouterr().out) is not suppressed
+
+
 def test_prometheus_config_parsing():
     """Test that prometheus metrics configuration is parsed correctly"""
     # Clear registry before test

tests/unit/proxy/db/test_check_migration.py

@@ -81,7 +81,7 @@ def test_a_skipped_migration_diff_is_one_json_record_in_json_mode(
     monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str]
 ) -> None:
     """
-    With json_logs on, the notice that the migration diff was skipped is one INFO record on stdout,
+    With json_logs on, the notice that the migration diff was skipped is one WARNING record on stdout,
     also while the loggers sit at WARNING, as they do at boot without LITELLM_LOG.
     """
     from litellm.proxy.db.check_migration import check_prisma_schema_diff_helper
@@ -101,7 +101,7 @@ def test_a_skipped_migration_diff_is_one_json_record_in_json_mode(
     assert len(lines) == 1, lines
     record: Final = json.loads(lines[0])
     assert (record["level"], record["message"]) == (
-        "INFO",
+        "WARNING",
         "Skipping the migration diff: litellm-proxy-extras has no Prisma runner. "
         "Error: import of litellm_proxy_extras.prisma_toolchain halted; None in sys.modules",
     )

tests/unit/proxy/policy_engine/test_init_policies.py

@@ -1,8 +1,10 @@
 """The policy table printed at boot keeps its terminal form in plain mode and is JSON when json_logs is on."""
 
 import json
+import re
 import sys
 from collections.abc import Mapping, Sequence
+from io import StringIO
 from typing import Final
 
 import pytest
@@ -40,6 +42,15 @@ NO_ATTACHMENTS_WARNING: Final = (
 )
 
 
+class _Stream(StringIO):
+    def __init__(self, terminal: bool) -> None:
+        super().__init__()
+        self._terminal = terminal
+
+    def isatty(self) -> bool:
+        return self._terminal
+
+
 @pytest.mark.parametrize(
     ("attachments", "expected"),
     [
@@ -53,18 +64,24 @@ NO_ATTACHMENTS_WARNING: Final = (
     ],
     ids=["with-attachments", "without-attachments"],
 )
+@pytest.mark.parametrize("terminal", [True, False], ids=["terminal", "piped"])
 def test_policy_table_prints_unchanged_in_plain_mode(
     monkeypatch: pytest.MonkeyPatch,
-    capfd: pytest.CaptureFixture[str],
     attachments: Sequence[Mapping[str, object]] | None,
     expected: str,
+    terminal: bool,
 ) -> None:
-    """Operators without JSON logs keep the colored policy table byte for byte."""
+    """Operators without JSON logs keep the colored policy table byte for byte on a terminal, and the same text
+    without color codes on a piped stdout."""
     monkeypatch.setattr(litellm, "json_logs", False)
+    monkeypatch.delenv("NO_COLOR", raising=False)
+    stdout: Final = _Stream(terminal)
+    monkeypatch.setattr(sys, "stdout", stdout)
+    monkeypatch.setattr(sys, "stderr", _Stream(terminal))
 
     _print_policies_on_startup(POLICIES, attachments)
 
-    assert capfd.readouterr() == (expected, "")
+    assert stdout.getvalue() == (expected if terminal else re.sub(r"\x1b\[\d+m", "", expected))
 
 
 def test_policy_table_is_one_json_record_in_json_mode(
@@ -98,8 +115,8 @@ def test_policy_table_is_one_json_record_in_json_mode(
 def test_missing_attachments_warning_is_a_second_json_record(
     monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str]
 ) -> None:
-    """Without attachments the table and the yellow warning are two INFO records on stdout, where the two
-    prints put them."""
+    """Without attachments the table and the yellow warning are an INFO and a WARNING record on stdout,
+    where the two prints put them."""
     monkeypatch.setattr(litellm, "json_logs", True)
     monkeypatch.setattr(sys, "excepthook", sys.excepthook)
     turn_on_json()
@@ -113,4 +130,4 @@ def test_missing_attachments_warning_is_a_second_json_record(
         "\n".join(("LiteLLM Policy Engine: Loaded 1 policies", "", *POLICY_ROWS)),
         NO_ATTACHMENTS_WARNING,
     ]
-    assert {record["level"] for record in records} == {"INFO"}
+    assert [record["level"] for record in records] == ["INFO", "WARNING"]

tests/unit/proxy/proxy_server/test_proxy_config.py

@@ -13,6 +13,7 @@ import json
 import logging
 import os
 import re
+import sys
 from collections.abc import Awaitable, Callable, Mapping
 from contextlib import nullcontext
 from dataclasses import dataclass
@@ -28,6 +29,7 @@ from prisma.errors import DataError
 from pydantic import JsonValue, TypeAdapter, ValidationError
 
 import litellm
+from litellm._logging import turn_on_json
 from litellm.proxy._types import CommonProxyErrors, ConfigGeneralSettings
 from litellm.proxy.common_utils.encrypt_decrypt_utils import encrypt_value_helper
 from litellm.types.router import Deployment, LiteLLM_Params
@@ -2088,6 +2090,118 @@ def test_ProxyConfig_parse_search_tools_missing_returns_none():
     assert pc.parse_search_tools({}) is None
 
 
+@pytest.mark.asyncio
+async def test_ProxyConfig_lists_search_tools_once_at_startup_not_on_every_reparse(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str]
+):
+    """The search tools reload job and add_deployment re-parse search_tools every config reload interval, so
+    the startup listing belongs to load_config alone, or a container log repeats it every 30 seconds."""
+    config_file: Final = tmp_path / "search.yaml"
+    config_file.write_text(
+        "model_list: []\n"
+        "search_tools:\n"
+        "  - search_tool_name: web-search\n"
+        "    litellm_params:\n"
+        "      search_provider: tavily\n"
+    )
+    monkeypatch.setattr(litellm, "json_logs", False)
+    monkeypatch.setattr(litellm, "suppress_startup_output", False)
+    monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", None)
+    monkeypatch.setattr("litellm.proxy.proxy_server.store_model_in_db", False)
+    monkeypatch.delenv("LITELLM_CONFIG_BUCKET_NAME", raising=False)
+    proxy_config: Final = ProxyConfig()
+
+    await proxy_config.load_config(router=None, config_file_path=str(config_file))
+    for _ in range(3):
+        assert proxy_config.parse_search_tools(proxy_config.get_config_state()) is not None
+
+    out: Final = capfd.readouterr().out
+    assert out.count("Proxy initialized with Search Tools:") == 1, out
+    assert out.count("web-search (tavily)") == 1, out
+
+
+@pytest.mark.asyncio
+async def test_ProxyConfig_lists_each_search_tool_as_its_own_json_record(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str]
+):
+    """With json_logs on, the startup listing is a header record and one record per tool, so a pipeline that
+    reads one tool per record keeps working."""
+    config_file: Final = tmp_path / "search.yaml"
+    config_file.write_text(
+        "model_list: []\n"
+        "search_tools:\n"
+        "  - search_tool_name: web-search\n"
+        "    litellm_params:\n"
+        "      search_provider: tavily\n"
+        "  - search_tool_name: news-search\n"
+        "    litellm_params:\n"
+        "      search_provider: exa_ai\n"
+    )
+    monkeypatch.setattr(litellm, "json_logs", True)
+    monkeypatch.setattr(litellm, "suppress_startup_output", False)
+    monkeypatch.setattr(sys, "excepthook", sys.excepthook)
+    monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", None)
+    monkeypatch.setattr("litellm.proxy.proxy_server.store_model_in_db", False)
+    monkeypatch.delenv("LITELLM_CONFIG_BUCKET_NAME", raising=False)
+    turn_on_json()
+
+    await ProxyConfig().load_config(router=None, config_file_path=str(config_file))
+
+    messages: Final = [json.loads(raw)["message"] for raw in capfd.readouterr().out.splitlines() if raw.strip()]
+    listing: Final = [message for message in messages if "Search Tools" in message or "-search (" in message]
+    assert listing == [
+        "LiteLLM: Proxy initialized with Search Tools:",
+        "    web-search (tavily)",
+        "    news-search (exa_ai)",
+    ], messages
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize(
+    ("already_on", "config_value"),
+    [
+        pytest.param(True, False, id="env-var-on-and-config-false"),
+        pytest.param(False, True, id="config-key-below-cache"),
+    ],
+)
+async def test_ProxyConfig_load_config_keeps_startup_output_off_once_either_setting_turns_it_on(
+    tmp_path: Path,
+    monkeypatch: pytest.MonkeyPatch,
+    capfd: pytest.CaptureFixture[str],
+    already_on: bool,
+    config_value: bool,
+):
+    """Only `true` turns suppress_startup_output on, from the env var or the config key, and a config `false`
+    never turns the env var back off. The key applies before any other litellm_settings key, so a `cache: true`
+    above it prints nothing either."""
+    config_file: Final = tmp_path / "suppress.yaml"
+    config_file.write_text(
+        "model_list: []\n"
+        "litellm_settings:\n"
+        "  cache: true\n"
+        "  cache_params:\n"
+        "    type: local\n"
+        f"  suppress_startup_output: {str(config_value).lower()}\n"
+        "search_tools:\n"
+        "  - search_tool_name: web-search\n"
+        "    litellm_params:\n"
+        "      search_provider: tavily\n"
+    )
+    monkeypatch.setattr(litellm, "json_logs", False)
+    monkeypatch.setattr(litellm, "suppress_startup_output", already_on)
+    monkeypatch.setattr(litellm, "cache", None)
+    monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", None)
+    monkeypatch.setattr("litellm.proxy.proxy_server.store_model_in_db", False)
+    monkeypatch.delenv("LITELLM_CONFIG_BUCKET_NAME", raising=False)
+
+    await ProxyConfig().load_config(router=None, config_file_path=str(config_file))
+
+    out: Final = capfd.readouterr().out
+    assert litellm.suppress_startup_output is True
+    assert "Setting Cache on Proxy" not in out, out
+    assert "Proxy initialized with Search Tools:" not in out, out
+
+
+
 def test_ProxyConfig_merge_config_and_db_search_tools_returns_superset():
     config_tools = [
         {

tests/unit/proxy/test_max_budget_env_var.py

@@ -18,10 +18,7 @@ async def test_max_budget_string_converted_to_float():
     string. initialize() should convert it to float so the comparison
     `litellm.max_budget > 0` doesn't raise TypeError.
     """
-    with (
-        patch("litellm.proxy.common_utils.banner.show_banner"),
-        patch("litellm.proxy.proxy_server.generate_feedback_box"),
-    ):
+    with patch("litellm.proxy.proxy_server.generate_feedback_box"):
         from litellm.proxy.proxy_server import initialize
 
         original = litellm.max_budget
@@ -36,10 +33,7 @@ async def test_max_budget_string_converted_to_float():
 @pytest.mark.asyncio
 async def test_max_budget_float_stays_float():
     """max_budget as float should still work."""
-    with (
-        patch("litellm.proxy.common_utils.banner.show_banner"),
-        patch("litellm.proxy.proxy_server.generate_feedback_box"),
-    ):
+    with patch("litellm.proxy.proxy_server.generate_feedback_box"):
         from litellm.proxy.proxy_server import initialize
 
         original = litellm.max_budget

tests/unit/proxy/test_proxy_cli.py

@@ -215,6 +215,23 @@ class TestProxyInitializationHelpers:
         )
         assert args["timeout_worker_healthcheck"] == 30
 
+    def test_an_old_uvicorn_drops_worker_healthcheck_and_warns_with_startup_output_suppressed(
+        self, monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
+    ) -> None:
+        class _UvicornConfigWithoutWorkerHealthcheck:
+            def __init__(self, app=None):
+                pass
+
+        monkeypatch.setattr(litellm, "suppress_startup_output", True)
+        monkeypatch.setattr(litellm, "json_logs", False)
+        with patch("uvicorn.Config", _UvicornConfigWithoutWorkerHealthcheck):
+            args: Final = ProxyInitializationHelpers._get_default_unvicorn_init_args(
+                "localhost", 8000, timeout_worker_healthcheck=15
+            )
+
+        assert "timeout_worker_healthcheck" not in args
+        assert "--timeout_worker_healthcheck requires uvicorn>=0.37.0" in capsys.readouterr().out
+
     def test_get_reload_options_no_config_still_watches_env(self):
         opts = ProxyInitializationHelpers._get_reload_options(None)
         assert opts["reload"] is True
@@ -1835,10 +1852,9 @@ class TestProxyInitializationHelpers:
         assert "uvicorn exits" not in warning
 
     @patch("uvicorn.run")
-    @patch("builtins.print")
     @patch("litellm.proxy.db.prisma_client.PrismaManager.setup_database")
     def test_single_worker_max_requests_warning_printed_on_uvicorn(
-        self, mock_setup_db, mock_print, mock_uvicorn_run
+        self, mock_setup_db, mock_uvicorn_run
     ):
         """uvicorn single-worker startup prints the does-not-recycle warning"""
         from click.testing import CliRunner
@@ -1881,11 +1897,8 @@ class TestProxyInitializationHelpers:
             assert (
                 result.exit_code == 0
             ), f"exit_code={result.exit_code}, output={result.output}"
-            assert any(
-                "--max_requests_before_restart 123 with --num_workers 1" in str(c)
-                and "uvicorn exits the process" in str(c)
-                for c in mock_print.call_args_list
-            )
+            assert "--max_requests_before_restart 123 with --num_workers 1" in result.output
+            assert "uvicorn exits the process" in result.output
 
     @patch("litellm.proxy.proxy_cli.start_query_engine_reaper")
     @patch("uvicorn.run")
@@ -1942,10 +1955,9 @@ class TestProxyInitializationHelpers:
 
     @patch("litellm.proxy.proxy_cli.ProxyInitializationHelpers.run_gunicorn_server")
     @patch("uvicorn.run")
-    @patch("builtins.print")
     @patch("litellm.proxy.db.prisma_client.PrismaManager.setup_database")
     def test_single_worker_max_requests_warning_printed_on_gunicorn(
-        self, mock_setup_db, mock_print, mock_uvicorn_run, mock_run_gunicorn
+        self, mock_setup_db, mock_uvicorn_run, mock_run_gunicorn
     ):
         """--run_gunicorn single-worker startup prints the gunicorn warning variant"""
         from click.testing import CliRunner
@@ -1989,11 +2001,8 @@ class TestProxyInitializationHelpers:
             assert (
                 result.exit_code == 0
             ), f"exit_code={result.exit_code}, output={result.output}"
-            assert any(
-                "--max_requests_before_restart 123 with --num_workers 1" in str(c)
-                and "gunicorn accepts no connections" in str(c)
-                for c in mock_print.call_args_list
-            )
+            assert "--max_requests_before_restart 123 with --num_workers 1" in result.output
+            assert "gunicorn accepts no connections" in result.output
 
     @patch("litellm.proxy.proxy_cli.ProxyInitializationHelpers.init_hypercorn_server")
     @patch("uvicorn.run")
@@ -2149,7 +2158,7 @@ class TestProxyInitializationHelpers:
         assert [call.args for call in reserved_before_run] == [("the gunicorn master",)]
 
     @pytest.mark.skipif(os.name == "nt", reason="gunicorn server path skips Windows")
-    def test_gunicorn_jitter_without_base_warns(self):
+    def test_gunicorn_jitter_without_base_warns(self, capsys: pytest.CaptureFixture[str]):
         """gunicorn path warns when jitter is set without --max_requests_before_restart"""
         pytest.importorskip("gunicorn")
 
@@ -2160,7 +2169,6 @@ class TestProxyInitializationHelpers:
 
         with (
             patch("gunicorn.app.base.BaseApplication.run", capture_run),
-            patch("builtins.print") as mock_print,
             patch(  # test-quality-ok: option tests must not start a thread or change the pytest worker's child ownership
                 "litellm.proxy.proxy_cli.start_query_engine_reaper"
             ),
@@ -2178,7 +2186,7 @@ class TestProxyInitializationHelpers:
 
         assert "max_requests" not in captured["options"]
         assert "max_requests_jitter" not in captured["options"]
-        assert any("has no effect" in str(c) for c in mock_print.call_args_list)
+        assert "has no effect" in capsys.readouterr().out
 
     def test_apply_uvicorn_jitter_sets_arg_when_supported(self):
         class _NewUvicornConfig:
@@ -2194,16 +2202,13 @@ class TestProxyInitializationHelpers:
             )
         assert uvicorn_args["limit_max_requests_jitter"] == 50
 
-    def test_apply_uvicorn_jitter_skipped_on_old_uvicorn(self):
+    def test_apply_uvicorn_jitter_skipped_on_old_uvicorn(self, capsys: pytest.CaptureFixture[str]):
         class _FakeUvicornConfig:
             def __init__(self, limit_max_requests=None):
                 pass
 
         uvicorn_args: dict = {}
-        with (
-            patch("uvicorn.Config", _FakeUvicornConfig),
-            patch("builtins.print") as mock_print,
-        ):
+        with patch("uvicorn.Config", _FakeUvicornConfig):
             ProxyInitializationHelpers._apply_uvicorn_max_requests_jitter(
                 uvicorn_args=uvicorn_args,
                 max_requests_before_restart=1000,
@@ -2211,19 +2216,18 @@ class TestProxyInitializationHelpers:
             )
 
         assert "limit_max_requests_jitter" not in uvicorn_args
-        assert any("0.41.0" in str(c) for c in mock_print.call_args_list)
+        assert "0.41.0" in capsys.readouterr().out
 
-    def test_apply_uvicorn_jitter_without_base_warns(self):
+    def test_apply_uvicorn_jitter_without_base_warns(self, capsys: pytest.CaptureFixture[str]):
         uvicorn_args: dict = {}
-        with patch("builtins.print") as mock_print:
-            ProxyInitializationHelpers._apply_uvicorn_max_requests_jitter(
-                uvicorn_args=uvicorn_args,
-                max_requests_before_restart=None,
-                jitter=50,
-            )
+        ProxyInitializationHelpers._apply_uvicorn_max_requests_jitter(
+            uvicorn_args=uvicorn_args,
+            max_requests_before_restart=None,
+            jitter=50,
+        )
 
         assert "limit_max_requests_jitter" not in uvicorn_args
-        assert any("has no effect" in str(c) for c in mock_print.call_args_list)
+        assert "has no effect" in capsys.readouterr().out
 
     @patch.dict(os.environ, {}, clear=True)
     def test_construct_database_url_from_env_vars(self):
@@ -2648,10 +2652,56 @@ class TestRunServerDbSetup:
             assert exit_info.value.code == 1
             assert "a database URL is set but the prisma CLI is neither on PATH nor importable" in out
             assert "pip install 'litellm[extra_proxy]'" in out
+            assert "\x1b[" not in out
         else:
             assert "prisma CLI" not in out
             assert "Setup complete" in out
 
+    @patch("atexit.register")
+    @patch("litellm.proxy.db.prisma_client.PrismaManager.setup_database", return_value=True)
+    @patch("litellm.proxy.db.prisma_client.PrismaManager.build_request_log_indexes", return_value=True)
+    def test_a_pgbouncer_that_cannot_start_stops_the_cli_with_an_uncolored_error_off_a_terminal(
+        self,
+        mock_build_indexes,
+        mock_setup_database,
+        mock_atexit_register,
+        tmp_path: Path,
+        capsys: pytest.CaptureFixture[str],
+    ) -> None:
+        from litellm.proxy.proxy_server import ProxyConfig
+
+        mock_proxy_module: Final = MagicMock(
+            app=MagicMock(),
+            ProxyConfig=ProxyConfig,
+            KeyManagementSettings=MagicMock(),
+            save_worker_config=MagicMock(),
+        )
+        clean_env: Final = {
+            k: v
+            for k, v in os.environ.items()
+            if not k.startswith(("DATABASE_URL", "DIRECT_URL", "LITELLM_PGBOUNCER_", "DISABLE_SCHEMA_UPDATE"))
+        }
+        pgbouncer_env: Final = {
+            "DATABASE_URL": "postgresql://test:test@localhost:5432/test",
+            "LITELLM_PGBOUNCER_ENABLED": "true",
+            "LITELLM_PGBOUNCER_BINARY": str(tmp_path / "missing-pgbouncer"),
+        }
+
+        with (
+            patch.dict(os.environ, {**clean_env, **pgbouncer_env}, clear=True),
+            patch.dict(
+                "sys.modules",
+                {"proxy_server": mock_proxy_module, "litellm.proxy.proxy_server": mock_proxy_module},
+            ),
+            pytest.raises(SystemExit) as exit_info,
+        ):
+            run_server.main(["--local", "--skip_server_startup"], standalone_mode=False)
+
+        error: Final = capsys.readouterr().err
+        assert exit_info.value.code == 1
+        assert "LITELLM_PGBOUNCER_ENABLED is set but the in-container pgbouncer could not start" in error
+        assert "\x1b[" not in error
+
     @patch("subprocess.run")
     @patch("atexit.register")
     @patch("litellm.proxy.db.prisma_client.PrismaManager.setup_database")
@@ -2726,6 +2776,7 @@ class TestRunServerDbSetup:
         mock_setup_database,
         mock_atexit_register,
         mock_subprocess_run,
+        capsys,
     ):
         """A sqlite DATABASE_URL must exit immediately, before any prisma call,
         instead of stalling on a migration against the postgresql-only schema."""
@@ -2764,6 +2815,9 @@ class TestRunServerDbSetup:
                 )
             assert exc_info.value.code == 1
             mock_setup_database.assert_not_called()
+        err: Final = capsys.readouterr().err
+        assert "sqlite" in err
+        assert "\x1b[" not in err
 
     @patch("subprocess.run")
     @patch("atexit.register")
@@ -2825,6 +2879,51 @@ class TestRunServerDbSetup:
         mock_setup_database.assert_called_once_with(use_migrate=True, use_v2_resolver=True)
         assert "--use_v2_migration_resolver is deprecated" not in capsys.readouterr().out
 
+    @patch("subprocess.run", return_value=MagicMock(returncode=0))
+    @patch("atexit.register")
+    @patch("litellm.proxy.db.prisma_client.PrismaManager.setup_database", return_value=True)
+    @patch("litellm.proxy.db.prisma_client.PrismaManager.build_request_log_indexes")
+    def test_the_deprecated_v2_resolver_flag_warns_with_startup_output_suppressed(
+        self,
+        mock_build_indexes,
+        mock_setup_database,
+        mock_atexit_register,
+        mock_subprocess_run,
+        monkeypatch: pytest.MonkeyPatch,
+        capsys: pytest.CaptureFixture[str],
+    ) -> None:
+        from litellm.proxy.proxy_server import ProxyConfig
+
+        mock_proxy_module: Final = MagicMock(
+            app=MagicMock(),
+            ProxyConfig=ProxyConfig,
+            KeyManagementSettings=MagicMock(),
+            save_worker_config=MagicMock(),
+        )
+        clean_env: Final = {
+            k: v
+            for k, v in os.environ.items()
+            if k not in ("DATABASE_URL", "DIRECT_URL", "USE_V2_MIGRATION_RESOLVER", "DISABLE_SCHEMA_UPDATE")
+        }
+        monkeypatch.setattr(litellm, "suppress_startup_output", True)
+        monkeypatch.setattr(litellm, "json_logs", False)
+
+        with (
+            patch.dict(
+                os.environ, {**clean_env, "DATABASE_URL": "postgresql://test:test@localhost:5432/test"}, clear=True
+            ),
+            patch.dict(
+                "sys.modules",
+                {"proxy_server": mock_proxy_module, "litellm.proxy.proxy_server": mock_proxy_module},
+            ),
+        ):
+            run_server.main(
+                ["--local", "--skip_server_startup", "--use_v2_migration_resolver"], standalone_mode=False
+            )
+
+        mock_setup_database.assert_called_once_with(use_migrate=True, use_v2_resolver=True)
+        assert "--use_v2_migration_resolver is deprecated and has no effect" in capsys.readouterr().out
+
     @pytest.mark.parametrize(
         ("arguments", "environment", "warned"),
         (
@@ -3968,11 +4067,104 @@ def _boot_from_config_file(config_path: Path, *flags: str) -> str | None:
         return os.environ.get("JSON_LOGS")
 
 
+COUNTING_SECRET_MANAGER_MODULE: Final = """
+import os
+
+from litellm.integrations.custom_secret_manager import CustomSecretManager
+
+
+class CountingSecretManager(CustomSecretManager):
+    def __init__(self):
+        super().__init__()
+        self._log("constructed")
+
+    @staticmethod
+    def _log(event):
+        with open(os.environ["SECRET_READ_LOG"], "a") as f:
+            f.write(event + "\\n")
+
+    def sync_read_secret(self, secret_name, optional_params=None, timeout=None, **kwargs):
+        self._log(secret_name)
+        return "from-secret-manager"
+
+    async def async_read_secret(self, secret_name, optional_params=None, timeout=None, **kwargs):
+        self._log(secret_name)
+        return "from-secret-manager"
+"""
+
+SECRET_MANAGER_BACKED_CONFIG: Final = """
+model_list:
+  - model_name: my-model
+    litellm_params:
+      model: openai/gpt-5.4-mini
+      api_key: os.environ/MY_PROVIDER_KEY
+
+litellm_settings:
+  suppress_startup_output: true
+
+general_settings:
+  key_management_system: custom
+  key_management_settings:
+    custom_secret_manager: counting_secret_manager.CountingSecretManager
+    hosted_keys:
+      - MY_PROVIDER_KEY
+"""
+
+
 class TestJsonLogsConfigReachesSpawnedWorkers:
     """`litellm_settings.json_logs: true` turns JSON on inside the CLI process, but the workers that
     uvicorn or granian spawn are fresh interpreters that pick plain or JSON from the JSON_LOGS
     environment variable at import. The CLI has to export it, or every worker logs plain text until
-    its own load_config reaches the key."""
+    its own load_config reaches the key. `suppress_startup_output` has the same problem with the banner
+    each worker prints before its load_config."""
+
+    @staticmethod
+    def _write_config(path: Path, litellm_settings: dict[str, object]) -> Path:
+        path.write_text(yaml.dump({"model_list": [], "litellm_settings": litellm_settings}))
+        return path
+
+    @staticmethod
+    def _env_after_boot(
+        name: str,
+        config: Path | None = None,
+        config_file_path_env: Path | None = None,
+        extra_env: dict[str, str] | None = None,
+        proxy_config_class: object = None,
+    ) -> str | None:
+        from litellm.proxy.proxy_server import ProxyConfig
+
+        mock_proxy_module: Final = MagicMock(
+            app=MagicMock(),
+            ProxyConfig=proxy_config_class or ProxyConfig,
+            KeyManagementSettings=MagicMock(),
+            save_worker_config=MagicMock(),
+        )
+        clean_env: Final = {
+            k: v
+            for k, v in os.environ.items()
+            if k
+            not in (
+                "DATABASE_URL",
+                "DIRECT_URL",
+                "JSON_LOGS",
+                "LITELLM_SUPPRESS_STARTUP_OUTPUT",
+                "CONFIG_FILE_PATH",
+                "LITELLM_CONFIG_BUCKET_NAME",
+            )
+        }
+        config_env: Final = {} if config_file_path_env is None else {"CONFIG_FILE_PATH": str(config_file_path_env)}
+        config_args: Final = [] if config is None else ["--config", str(config)]
+        with (
+            patch.dict(os.environ, {**clean_env, **config_env, **(extra_env or {})}, clear=True),
+            patch.dict(
+                "sys.modules",
+                {"proxy_server": mock_proxy_module, "litellm.proxy.proxy_server": mock_proxy_module},
+            ),
+            patch("atexit.register"),
+            patch("litellm.turn_on_json"),  # test-quality-ok: it swaps every logger handler in this pytest worker
+        ):
+            run_server.main([*config_args, "--local", "--skip_server_startup"], standalone_mode=False)
+            return os.environ.get(name)
 
     def test_json_logs_true_exports_json_logs_to_spawned_workers(
         self, tmp_path: Path, monkeypatch: pytest.MonkeyPatch
@@ -3991,11 +4183,126 @@ class TestJsonLogsConfigReachesSpawnedWorkers:
 
         assert _boot_from_config_file(config_path) is None
 
+    @pytest.mark.parametrize("litellm_settings", [[], "", 0, False])
+    @pytest.mark.parametrize("via_config_file_path_env", [False, True])
+    def test_an_empty_litellm_settings_still_boots(
+        self, tmp_path: Path, monkeypatch: pytest.MonkeyPatch, litellm_settings: object, via_config_file_path_env: bool
+    ) -> None:
+        monkeypatch.setattr(litellm, "suppress_startup_output", False)
+        config_file: Final = tmp_path / "config.yaml"
+        config_file.write_text(yaml.dump({"model_list": [], "litellm_settings": litellm_settings}))
+        exported: Final = (
+            self._env_after_boot("LITELLM_SUPPRESS_STARTUP_OUTPUT", config_file_path_env=config_file)
+            if via_config_file_path_env
+            else self._env_after_boot("LITELLM_SUPPRESS_STARTUP_OUTPUT", config=config_file)
+        )
+        assert exported is None
+
+    @pytest.mark.parametrize(("setting", "exported"), [(True, "true"), ("true", None), (False, None)])
+    def test_only_a_boolean_suppress_startup_output_reaches_spawned_workers(
+        self, tmp_path: Path, monkeypatch: pytest.MonkeyPatch, setting: object, exported: str | None
+    ) -> None:
+        monkeypatch.setattr(litellm, "suppress_startup_output", False)
+        config: Final = self._write_config(tmp_path / "config.yaml", {"suppress_startup_output": setting})
+        assert self._env_after_boot("LITELLM_SUPPRESS_STARTUP_OUTPUT", config=config) == exported
+
+    @pytest.mark.parametrize(("name", "exported"), [("LITELLM_SUPPRESS_STARTUP_OUTPUT", "true"), ("JSON_LOGS", None)])
+    def test_a_config_named_by_CONFIG_FILE_PATH_passes_only_suppress_startup_output_to_spawned_workers(
+        self, tmp_path: Path, monkeypatch: pytest.MonkeyPatch, name: str, exported: str | None
+    ) -> None:
+        """Its json_logs stays the workers' own to read, as before, so uvicorn's lines keep their plain format."""
+        monkeypatch.setattr(litellm, "json_logs", False)
+        monkeypatch.setattr(litellm, "suppress_startup_output", False)
+        settings: Final[dict[str, object]] = {"json_logs": True, "suppress_startup_output": True}
+        config_file: Final = self._write_config(tmp_path / "config.yaml", settings)
+        assert self._env_after_boot(name, config_file_path_env=config_file) == exported
+
+    def test_a_CONFIG_FILE_PATH_quiet_config_wins_over_a_cli_config_too(
+        self, tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+    ) -> None:
+        """Workers load CONFIG_FILE_PATH ahead of --config, so its key quiets the CLI when both are set."""
+        monkeypatch.setattr(litellm, "suppress_startup_output", False)
+        cli_config: Final = self._write_config(tmp_path / "cli.yaml", {})
+        env_config: Final = self._write_config(tmp_path / "env.yaml", {"suppress_startup_output": True})
+        exported: Final = self._env_after_boot(
+            "LITELLM_SUPPRESS_STARTUP_OUTPUT", config=cli_config, config_file_path_env=env_config
+        )
+        assert exported == "true"
+
+    @pytest.mark.parametrize(
+        ("config_file", "env_file", "extra_env"),
+        [
+            ("config.yaml", "missing.yaml", {}),
+            ("config.json", "config.json", {}),
+            ("config.yaml", "config.yaml", {"LITELLM_CONFIG_BUCKET_NAME": "litellm-configs"}),
+        ],
+        ids=["missing-file", "not-yaml", "bucket-config"],
+    )
+    def test_a_CONFIG_FILE_PATH_workers_skip_is_skipped_by_the_cli_too(
+        self,
+        tmp_path: Path,
+        monkeypatch: pytest.MonkeyPatch,
+        config_file: str,
+        env_file: str,
+        extra_env: dict[str, str],
+    ) -> None:
+        """Workers load CONFIG_FILE_PATH only when it names a YAML file and no config bucket is set, and a boot
+        with any other value starts, so the CLI reads nothing from it either instead of failing the boot."""
+        monkeypatch.setattr(litellm, "suppress_startup_output", False)
+        self._write_config(tmp_path / config_file, {"suppress_startup_output": True})
+        exported: Final = self._env_after_boot(
+            "LITELLM_SUPPRESS_STARTUP_OUTPUT", config_file_path_env=tmp_path / env_file, extra_env=extra_env
+        )
+        assert exported is None
+
+    def test_a_CONFIG_FILE_PATH_boot_reads_no_secret_before_the_workers_do(
+        self, tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+    ) -> None:
+        """The CLI takes only the suppress key from CONFIG_FILE_PATH, so a config backed by a secret manager is
+        read from it once per worker, as before, not once more by the CLI."""
+        from litellm.proxy.proxy_server import ProxyConfig
+
+        monkeypatch.setattr(litellm, "suppress_startup_output", False)
+        monkeypatch.setattr(litellm, "secret_manager_client", None)
+        monkeypatch.setattr(litellm, "_key_management_system", None)
+        monkeypatch.setattr(litellm, "_key_management_settings", litellm._key_management_settings)
+        reads: Final = tmp_path / "secret_reads.log"
+        (tmp_path / "counting_secret_manager.py").write_text(COUNTING_SECRET_MANAGER_MODULE)
+        config_file: Final = tmp_path / "config.yaml"
+        config_file.write_text(SECRET_MANAGER_BACKED_CONFIG)
+
+        exported: Final = self._env_after_boot(
+            "LITELLM_SUPPRESS_STARTUP_OUTPUT",
+            config_file_path_env=config_file,
+            extra_env={"SECRET_READ_LOG": str(reads)},
+            proxy_config_class=ProxyConfig,
+        )
+
+        assert exported == "true"
+        assert not reads.exists(), reads.read_text()
+
+    def test_a_CONFIG_FILE_PATH_key_set_in_an_included_file_reaches_spawned_workers(
+        self, tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+    ) -> None:
+        from litellm.proxy.proxy_server import ProxyConfig
+
+        monkeypatch.setattr(litellm, "suppress_startup_output", False)
+        self._write_config(tmp_path / "quiet.yaml", {"suppress_startup_output": True})
+        config_file: Final = tmp_path / "config.yaml"
+        config_file.write_text(yaml.dump({"include": ["quiet.yaml"], "model_list": []}))
+
+        exported: Final = self._env_after_boot(
+            "LITELLM_SUPPRESS_STARTUP_OUTPUT", config_file_path_env=config_file, proxy_config_class=ProxyConfig
+        )
+
+        assert exported == "true"
+
 
 class TestEnforcePrismaMigrationCheckNoticeAroundTheConfigRead:
     """With a config file, the `--enforce_prisma_migration_check` deprecation notice waits until the config
-    is read, so a `json_logs: true` set only there makes it a JSON record. It is printed once either way,
-    and a boot whose config read fails still prints it before the error, as it did before the wait."""
+    is read, so a `json_logs: true` set only there makes it a JSON record, while `suppress_startup_output: true`
+    leaves it, since it is a warning. Otherwise it is printed once, and a boot whose config read fails still
+    prints it before the error, as it did before the wait."""
 
     NOTICE: Final = "--enforce_prisma_migration_check is deprecated and has no effect"
     FLAG: Final = "--enforce_prisma_migration_check"
@@ -4035,13 +4342,26 @@ class TestEnforcePrismaMigrationCheckNoticeAroundTheConfigRead:
         assert len(notice_lines) == 1, notice_lines
         assert self.NOTICE in json.loads(notice_lines[0])["message"]
 
+    @pytest.mark.parametrize("json_logs", [False, True])
+    def test_a_config_only_suppress_startup_output_keeps_the_notice(
+        self, tmp_path: Path, monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str], json_logs: bool
+    ) -> None:
+        monkeypatch.setattr(litellm, "json_logs", json_logs)
+        monkeypatch.setattr(litellm, "suppress_startup_output", False)
+        config_path: Final = tmp_path / "config.yaml"
+        config_path.write_text(yaml.dump({"model_list": [], "litellm_settings": {"suppress_startup_output": True}}))
+
+        _boot_from_config_file(config_path, self.FLAG)
+
+        assert capsys.readouterr().out.count(self.NOTICE) == 1
+
 
 def test_granian_flag_notices_are_json_records_in_json_mode(
     monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str]
 ) -> None:
     """The yellow notices for the flags Granian ignores (`--max_requests_before_restart`, `--ciphers`) are
-    printed while the launcher starts; with json_logs on they are INFO records on stdout like every other
-    line, not plaintext between the records."""
+    printed while the launcher starts; with json_logs on they are WARNING records on stdout after the INFO
+    start line, not plaintext between the records."""
     pytest.importorskip("granian")
     monkeypatch.setattr(litellm, "json_logs", True)
     monkeypatch.setattr(sys, "excepthook", sys.excepthook)
@@ -4072,4 +4392,4 @@ def test_granian_flag_notices_are_json_records_in_json_mode(
         "(Granian uses workers_lifetime in seconds, not a per-request limit).",
         "LiteLLM: --ciphers is not applied when using --run_granian.",
     ]
-    assert {record["level"] for record in records} == {"INFO"}
+    assert [record["level"] for record in records] == ["INFO", "WARNING", "WARNING"]

tests/unit/proxy/test_proxy_server_endpoints_and_startup.py

@@ -15615,8 +15615,8 @@ def test_feedback_box_is_one_json_record_in_json_mode(
 async def test_initialize_prints_the_banner_between_blank_lines_in_plain_mode(
     monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str]
 ) -> None:
-    """Operators without JSON logs see the banner exactly as before, through show_banner()'s flushing
-    click.echo."""
+    """Operators without JSON logs see the banner exactly as before, flushed at once as show_banner()'s
+    click.echo did."""
     monkeypatch.setattr(litellm, "json_logs", False)
     monkeypatch.setenv("LITELLM_DONT_SHOW_FEEDBACK_BOX", "true")
 

tests/unit/test_logging.py

@@ -7,7 +7,7 @@ import logging
 import re
 import sys
 import time
-from io import StringIO
+from io import BytesIO, StringIO, TextIOWrapper
 from pathlib import Path
 from typing import Final, List
 
@@ -31,10 +31,11 @@ from litellm._logging import (
     StdoutLogTruncationFilter,
     _get_uvicorn_json_log_config,
     _initialize_loggers_with_handler,
-    _parse_json_logs_env,
+    _parse_env_flag,
     _plain_log_format,
     _stdout_truncation_marker,
     print_startup_message,
+    print_startup_warning,
     turn_on_json,
     format_base64_size,
     session_id_var,
@@ -46,6 +47,7 @@ from litellm._logging import (
     verbose_router_logger,
 )
 from litellm.constants import LITELLM_TRUNCATED_PAYLOAD_FIELD
+from litellm.proxy.common_utils.banner import LITELLM_BANNER
 from litellm.integrations.custom_logger import CustomLogger
 from litellm.litellm_core_utils import secret_redaction
 from litellm.types.utils import StandardLoggingPayload
@@ -1402,6 +1404,32 @@ def test_plain_log_format_survives_hostile_streams(stdout, stderr):
     assert _plain_log_format(stdout, stderr) == _PLAIN_LOG_FORMAT
 
 
+def _router_success_record() -> logging.LogRecord:
+    return logging.LogRecord(
+        "LiteLLM Router", logging.INFO, "router.py", 1, "litellm.acompletion(model=%s)\033[32m 200 OK\033[0m", ("gpt-5.6",), None
+    )
+
+
+@pytest.mark.parametrize(
+    ("log_format", "want_suffix"),
+    [
+        (_COLOR_LOG_FORMAT, "litellm.acompletion(model=gpt-5.6)\033[32m 200 OK\033[0m"),
+        (_PLAIN_LOG_FORMAT, "router.py:1 - litellm.acompletion(model=gpt-5.6) 200 OK"),
+    ],
+)
+def test_plain_formatter_keeps_message_colors_only_with_the_color_format(log_format: str, want_suffix: str):
+    """A message that embeds color codes keeps them on a color terminal. Piped, or under NO_COLOR, the plain
+    format is chosen and the codes go too, so a container log never stores them as text."""
+    assert CorrelationPlainFormatter(log_format).format(_router_success_record()).endswith(want_suffix)
+
+
+def test_json_formatter_drops_color_codes_from_the_message():
+    """A JSON log line is read by a log pipeline, never a terminal, so a colored message arrives as plain text."""
+    assert json.loads(JsonFormatter().format(_router_success_record()))["message"] == (
+        "litellm.acompletion(model=gpt-5.6) 200 OK"
+    )
+
+
 def test_level_routing_handler_falls_back_to_stderr_when_stdout_is_unusable(monkeypatch, capsys):
     logger = logging.getLogger("test_level_routing_fallback")
     logger.handlers.clear()
@@ -1436,9 +1464,9 @@ def test_level_routing_handler_falls_back_to_stderr_when_stdout_is_unusable(monk
         (None, False),
     ],
 )
-def test_parse_json_logs_env_enables_only_on_true(value, want):
+def test_parse_env_flag_enables_only_on_true(value: str | None, want: bool):
     """JSON_LOGS=false / 0 must not enable JSON logs (LIT-5558)."""
-    assert _parse_json_logs_env(value) is want
+    assert _parse_env_flag(value) is want
 
 
 def test_plain_log_format_survives_none_streams():
@@ -1813,17 +1841,98 @@ def test_print_verbose_formats_args_only_when_set_verbose_is_on(monkeypatch, cap
     assert capsys.readouterr().out == "key k1 value cached-vector\n"
 
 
-def test_print_startup_message_prints_plain_mode_text_unchanged(
-    monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str]
-) -> None:
-    """Operators without JSON logs keep the colored, spaced terminal output exactly as before."""
+class _CapturedStream(StringIO):
+    def __init__(self, tty: bool) -> None:
+        super().__init__()
+        self._tty = tty
+
+    def isatty(self) -> bool:
+        return self._tty
+
+
+@pytest.mark.parametrize(
+    "stdout_tty, stderr_tty, no_color, want_color",
+    [(True, True, None, True), (False, True, None, False), (True, False, None, False), (True, True, "1", False)],
+)
+def test_print_startup_message_keeps_colors_only_on_a_color_terminal(
+    monkeypatch: pytest.MonkeyPatch, stdout_tty: bool, stderr_tty: bool, no_color: str | None, want_color: bool
+):
+    """Operators without JSON logs keep the colored, spaced output on a terminal. A container runtime reads
+    stdout through a pipe and stores the escape codes as text, so there, and under NO_COLOR, the colors go.
+    The rule is the plain log format's, so a `2>file` run drops the colors from prints and log lines alike."""
     monkeypatch.setattr(litellm, "json_logs", False)
+    if no_color is None:
+        monkeypatch.delenv("NO_COLOR", raising=False)
+    else:
+        monkeypatch.setenv("NO_COLOR", no_color)
+    stdout: Final = _CapturedStream(stdout_tty)
+    monkeypatch.setattr(sys, "stdout", stdout)
+    monkeypatch.setattr(sys, "stderr", _CapturedStream(stderr_tty))
+    assert (_plain_log_format(sys.stdout, sys.stderr) == _COLOR_LOG_FORMAT) is want_color
 
     print_startup_message("\n\033[1;32mLiteLLM: Proxy initialized with Config, Set models:\033[0m\n")
 
+    assert stdout.getvalue() == (
+        "\n\033[1;32mLiteLLM: Proxy initialized with Config, Set models:\033[0m\n\n"
+        if want_color
+        else "\nLiteLLM: Proxy initialized with Config, Set models:\n\n"
+    )
+
+
+def test_print_startup_message_writes_the_banner_as_utf8_to_an_ascii_stdout(monkeypatch: pytest.MonkeyPatch):
+    monkeypatch.setattr(litellm, "json_logs", False)
+    monkeypatch.setattr(litellm, "suppress_startup_output", False)
+    raw: Final = BytesIO()
+    monkeypatch.setattr(sys, "stdout", TextIOWrapper(raw, encoding="ascii"))
+
+    print_startup_message(f"\n{LITELLM_BANNER}\n")
+
+    assert raw.getvalue().decode("utf-8") == f"\n{LITELLM_BANNER}\n\n"
+
+
+@pytest.mark.parametrize("json_logs", [False, True])
+def test_print_startup_message_writes_nothing_when_suppressed(
+    monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str], json_logs: bool
+):
+    """suppress_startup_output turns the startup messages off in plain and JSON mode alike, for operators
+    whose log pipeline wants only the logger's own lines."""
+    monkeypatch.setattr(litellm, "json_logs", json_logs)
+    monkeypatch.setattr(litellm, "suppress_startup_output", True)
+    monkeypatch.setattr(sys, "excepthook", sys.excepthook)
+    if json_logs:
+        turn_on_json()
+
+    print_startup_message("\033[32mLiteLLM: Proxy initialized with Search Tools:\033[0m\n\033[32m    web-search (tavily)\033[0m")
+
+    assert capfd.readouterr() == ("", "")
+
+
+@pytest.mark.parametrize("json_logs", [False, True])
+def test_print_startup_warning_still_prints_when_suppressed(
+    monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str], json_logs: bool
+):
+    """A warning names something the operator has to fix, like an ignored flag or a failed migration diff,
+    so suppress_startup_output leaves it in place. In JSON mode it is a WARNING record at the caller's
+    location, on stdout where the print put it, so a collector's WARNING filter keeps it."""
+    monkeypatch.setattr(litellm, "json_logs", json_logs)
+    monkeypatch.setattr(litellm, "suppress_startup_output", True)
+    monkeypatch.setattr(sys, "excepthook", sys.excepthook)
+    if json_logs:
+        turn_on_json()
+
+    print_startup_warning("\033[1;33mLiteLLM: --ciphers is not applied when using --run_granian.\033[0m\n")
+
     out, err = capfd.readouterr()
-    assert out == "\n\033[1;32mLiteLLM: Proxy initialized with Config, Set models:\033[0m\n\n"
     assert err == ""
+    lines: Final = [raw for raw in out.splitlines() if raw.strip()]
+    assert len(lines) == 1, lines
+    if json_logs:
+        record: Final = json.loads(lines[0])
+        assert record["message"] == "LiteLLM: --ciphers is not applied when using --run_granian."
+        assert record["level"] == "WARNING"
+        assert str(record["logger"]).startswith("test_logging.py:"), record["logger"]
+    else:
+        assert lines[0] == "LiteLLM: --ciphers is not applied when using --run_granian."
 
 
 def test_print_startup_message_is_one_json_record_in_json_mode(
@@ -1852,27 +1961,6 @@ def test_print_startup_message_is_one_json_record_in_json_mode(
     assert str(record["logger"]).startswith("test_logging.py:"), record["logger"]
 
 
-def test_print_startup_message_yellow_notice_is_an_info_record_on_stdout(
-    monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str]
-) -> None:
-    """An operator notice printed in yellow (an ignored flag, a deprecation) stays on stdout in JSON mode,
-    where the print put it, as one INFO record with the color codes removed."""
-    monkeypatch.setattr(litellm, "json_logs", True)
-    monkeypatch.setattr(sys, "excepthook", sys.excepthook)
-    turn_on_json()
-    verbose_proxy_logger.setLevel(logging.WARNING)
-
-    print_startup_message("\033[1;33mLiteLLM: --ciphers is not applied when using --run_granian.\033[0m\n")
-
-    out, err = capfd.readouterr()
-    assert err == ""
-    lines: Final = [raw for raw in out.splitlines() if raw.strip()]
-    assert len(lines) == 1, lines
-    record: Final = json.loads(lines[0])
-    assert record["message"] == "LiteLLM: --ciphers is not applied when using --run_granian."
-    assert record["level"] == "INFO"
-
-
 def test_print_startup_message_shows_above_the_json_handler_level(
     monkeypatch: pytest.MonkeyPatch, capfd: pytest.CaptureFixture[str]
 ) -> None: