fix(proxy): give every requests call a timeout so a silent server cannot hang the caller

`requests` has no default timeout, so a host that accepts the connection and
never answers blocks the calling thread forever.

The one on the request path is the HiddenLayer guardrail's `_get_jwt`. It runs
synchronously inside `_call_hiddenlayer` whenever the hour-long JWT expires and
the API answers 401, so a stalled auth host parked the worker's whole event
loop, not just the guarded request. The other eight are the teams and users CLI
clients, which pin the operator's terminal instead.

`TeamsManagementClient` and `UsersManagementClient` now take the same
`timeout: int = 30` their `HTTPClient` sibling already had, and `Client` threads
its own timeout down to teams. `_poll_for_ready_data` already passed a timeout
through a TypedDict that ruff could not see into; passing the argument directly
retires both the TypedDict and the suppression it would have needed.

Graduate S113 into ruff.toml so the next `requests` call without a timeout fails
the lint step.

litellm/proxy/client/cli/commands/auth.py

@@ -99,11 +99,6 @@ class CliPollData(TypedDict, total=False):
     team_id: str
 
 
-class CliPollRequestKwargs(TypedDict, total=False):
-    timeout: int
-    headers: dict[str, str]
-
-
 class CliSsoStartData(TypedDict):
     login_id: str
     poll_secret: str
@@ -518,10 +513,7 @@ def _poll_for_ready_data(
 ) -> CliPollData | None:
     for attempt in range(total_timeout // poll_interval):
         try:
-            request_kwargs: CliPollRequestKwargs = {"timeout": request_timeout}
-            if headers is not None:
-                request_kwargs["headers"] = headers
-            response = requests.get(url, **request_kwargs)
+            response = requests.get(url, headers=headers, timeout=request_timeout)
             if response.status_code == 200:
                 data: CliPollData = response.json()
                 status = data.get("status")

litellm/proxy/client/client.py

@@ -38,4 +38,4 @@ class Client:
         self.chat = ChatClient(base_url=self._base_url, api_key=self._api_key)
         self.keys = KeysManagementClient(base_url=self._base_url, api_key=self._api_key)
         self.credentials = CredentialsManagementClient(base_url=self._base_url, api_key=self._api_key)
-        self.teams = TeamsManagementClient(base_url=self._base_url, api_key=self._api_key)
+        self.teams = TeamsManagementClient(base_url=self._base_url, api_key=self._api_key, timeout=timeout)

litellm/proxy/client/teams.py

@@ -11,16 +11,18 @@ from .exceptions import UnauthorizedError
 class TeamsManagementClient:
     """Client for managing teams in LiteLLM proxy."""
 
-    def __init__(self, base_url: str, api_key: str | None = None):
+    def __init__(self, base_url: str, api_key: str | None = None, timeout: int = 30):
         """
         Initialize the TeamsManagementClient.
 
         Args:
             base_url (str): The base URL of the LiteLLM proxy server (e.g., "http://localhost:4000")
             api_key (Optional[str]): API key for authentication. If provided, it will be sent as a Bearer token.
+            timeout (int): Request timeout in seconds (default: 30)
         """
         self._base_url = base_url.rstrip("/")  # Remove trailing slash if present
         self._api_key = api_key
+        self._timeout = timeout
 
     def _get_headers(self) -> dict[str, str]:
         """
@@ -60,7 +62,7 @@ class TeamsManagementClient:
         if organization_id:
             params["organization_id"] = organization_id
 
-        response: Final = requests.get(url, headers=self._get_headers(), params=params)
+        response: Final = requests.get(url, headers=self._get_headers(), params=params, timeout=self._timeout)
 
         if response.status_code == 401:
             raise UnauthorizedError("Authentication failed. Check your API key.")
@@ -117,7 +119,7 @@ class TeamsManagementClient:
         if sort_by:
             params["sort_by"] = sort_by
 
-        response: Final = requests.get(url, headers=self._get_headers(), params=params)
+        response: Final = requests.get(url, headers=self._get_headers(), params=params, timeout=self._timeout)
 
         if response.status_code == 401:
             raise UnauthorizedError("Authentication failed. Check your API key.")
@@ -138,7 +140,7 @@ class TeamsManagementClient:
         """
         url: Final = f"{self._base_url}/team/available"
 
-        response: Final = requests.get(url, headers=self._get_headers())
+        response: Final = requests.get(url, headers=self._get_headers(), timeout=self._timeout)
 
         if response.status_code == 401:
             raise UnauthorizedError("Authentication failed. Check your API key.")

litellm/proxy/client/users.py

@@ -6,9 +6,10 @@ from .exceptions import NotFoundError, UnauthorizedError
 
 
 class UsersManagementClient:
-    def __init__(self, base_url: str, api_key: str | None = None):
+    def __init__(self, base_url: str, api_key: str | None = None, timeout: int = 30):
         self.base_url = base_url.rstrip("/")
         self.api_key = api_key
+        self.timeout = timeout
 
     def _get_headers(self) -> dict[str, str]:
         headers: Final = {"Content-Type": "application/json"}
@@ -19,7 +20,7 @@ class UsersManagementClient:
     def list_users(self, params: dict[str, Any] | None = None) -> list[dict[str, Any]]:
         """List users (GET /user/list)"""
         url: Final = f"{self.base_url}/user/list"
-        response: Final = requests.get(url, headers=self._get_headers(), params=params)
+        response: Final = requests.get(url, headers=self._get_headers(), params=params, timeout=self.timeout)
         if response.status_code == 401:
             raise UnauthorizedError(response.text)
         response.raise_for_status()
@@ -29,7 +30,7 @@ class UsersManagementClient:
         """Get user info (GET /user/info)"""
         url: Final = f"{self.base_url}/user/info"
         params: Final = {"user_id": user_id} if user_id else {}
-        response: Final = requests.get(url, headers=self._get_headers(), params=params)
+        response: Final = requests.get(url, headers=self._get_headers(), params=params, timeout=self.timeout)
         if response.status_code == 401:
             raise UnauthorizedError(response.text)
         if response.status_code == 404:
@@ -41,7 +42,7 @@ class UsersManagementClient:
         """Get user info v2 - lightweight, returns only user object (GET /v2/user/info)"""
         url: Final = f"{self.base_url}/v2/user/info"
         params: Final = {"user_id": user_id} if user_id else {}
-        response: Final = requests.get(url, headers=self._get_headers(), params=params)
+        response: Final = requests.get(url, headers=self._get_headers(), params=params, timeout=self.timeout)
         if response.status_code == 401:
             raise UnauthorizedError(response.text)
         if response.status_code == 404:
@@ -52,7 +53,7 @@ class UsersManagementClient:
     def create_user(self, user_data: dict[str, Any]) -> dict[str, Any]:
         """Create a new user (POST /user/new)"""
         url: Final = f"{self.base_url}/user/new"
-        response: Final = requests.post(url, headers=self._get_headers(), json=user_data)
+        response: Final = requests.post(url, headers=self._get_headers(), json=user_data, timeout=self.timeout)
         if response.status_code == 401:
             raise UnauthorizedError(response.text)
         response.raise_for_status()
@@ -61,7 +62,9 @@ class UsersManagementClient:
     def delete_user(self, user_ids: list[str]) -> dict[str, Any]:
         """Delete users (POST /user/delete)"""
         url: Final = f"{self.base_url}/user/delete"
-        response: Final = requests.post(url, headers=self._get_headers(), json={"user_ids": user_ids})
+        response: Final = requests.post(
+            url, headers=self._get_headers(), json={"user_ids": user_ids}, timeout=self.timeout
+        )
         if response.status_code == 401:
             raise UnauthorizedError(response.text)
         response.raise_for_status()

litellm/proxy/guardrails/guardrail_hooks/hiddenlayer/hiddenlayer.py

@@ -36,6 +36,9 @@ if TYPE_CHECKING:
     from litellm.types.proxy.guardrails.guardrail_hooks.base import GuardrailConfigModel
 
 
+_AUTH_TIMEOUT_SECONDS: Final[float] = 30.0
+
+
 class _HiddenlayerEvaluation(TypedDict, total=False):
     action: str
     threat_level: str
@@ -117,10 +120,10 @@ def is_saas(host: str) -> bool:
     return False
 
 
-def _get_jwt(auth_url, api_id, api_key) -> str:
+def _get_jwt(auth_url, api_id, api_key, timeout: float = _AUTH_TIMEOUT_SECONDS) -> str:
     token_url: Final = f"{auth_url}/oauth2/token?grant_type=client_credentials"
 
-    resp: Final = requests.post(token_url, auth=HTTPBasicAuth(api_id, api_key))
+    resp: Final = requests.post(token_url, auth=HTTPBasicAuth(api_id, api_key), timeout=timeout)
 
     if not resp.ok:
         raise RuntimeError(

ruff.toml

@@ -6,7 +6,8 @@ lint.extend-select = [
     "T20", "PGH004", "RUF008", "RUF009", "RUF100",
     "B033", "FURB136", "FURB168", "FURB188", "I001", "PERF402", "PIE790", "PIE800", "PLC0208",
     "PLR0402", "PLR1711", "PLR1730", "PLR2044", "PLW0133", "PYI030", "PYI041", "PYI064", "RET501",
-    "RUF010", "RUF022", "RUF023", "RUF051", "SIM114", "SIM118", "TC005", "UP006", "UP007", "UP008",
+    "RUF010", "RUF022", "RUF023", "RUF051", "S113", "SIM114", "SIM118", "TC005", "UP006", "UP007",
+    "UP008",
     "UP012", "UP018", "UP024", "UP032", "UP034", "UP035", "UP037", "UP045",
 ]
 # RUF100 (unused-noqa) only knows the rules enabled in THIS config, so it would strip

tests/test_litellm/proxy/client/conftest.py

@@ -0,0 +1,38 @@
+import threading
+
+import pytest
+
+
+@pytest.fixture
+def hanging_server():
+    """A server that accepts the connection and never answers, so only a timeout ends the call."""
+    from http.server import BaseHTTPRequestHandler, HTTPServer
+    from socketserver import ThreadingMixIn
+
+    stop: threading.Event = threading.Event()
+
+    class SilentRequestHandler(BaseHTTPRequestHandler):
+        protocol_version = "HTTP/1.1"
+
+        def _hang(self):
+            stop.wait(timeout=30)
+
+        do_GET = _hang
+        do_POST = _hang
+
+        def log_message(self, format, *args):
+            pass
+
+    class ThreadedServer(ThreadingMixIn, HTTPServer):
+        daemon_threads = True
+
+    server = ThreadedServer(("127.0.0.1", 0), SilentRequestHandler)
+    thread = threading.Thread(target=server.serve_forever, daemon=True)
+    thread.start()
+    try:
+        yield f"http://127.0.0.1:{server.server_port}"
+    finally:
+        stop.set()
+        server.shutdown()
+        server.server_close()
+        thread.join(timeout=5)

tests/test_litellm/proxy/client/test_teams.py

@@ -0,0 +1,20 @@
+import time
+
+import pytest
+import requests
+
+from litellm.proxy.client.teams import TeamsManagementClient
+
+
+def test_list_gives_up_at_the_timeout_instead_of_hanging(hanging_server):
+    """
+    A proxy that accepts the connection but never answers used to pin the caller's
+    process forever, since the request carried no timeout at all.
+    """
+    client = TeamsManagementClient(base_url=hanging_server, api_key="sk-test", timeout=1)
+
+    started = time.monotonic()
+    with pytest.raises(requests.exceptions.Timeout):
+        client.list()
+
+    assert time.monotonic() - started < 10

tests/test_litellm/proxy/client/test_users.py

@@ -1,6 +1,8 @@
+import time
 from unittest.mock import MagicMock, patch
 
 import pytest
+import requests
 
 
 
@@ -82,3 +84,17 @@ def test_delete_user_unauthorized(mock_post, client):
     mock_post.return_value.text = "unauthorized"
     with pytest.raises(UnauthorizedError):
         client.delete_user(["u1"])
+
+
+def test_delete_user_gives_up_at_the_timeout_instead_of_hanging(hanging_server):
+    """
+    A proxy that accepts the connection but never answers used to pin the caller's
+    process forever, since the request carried no timeout at all.
+    """
+    client = UsersManagementClient(base_url=hanging_server, api_key="sk-test", timeout=1)
+
+    started = time.monotonic()
+    with pytest.raises(requests.exceptions.Timeout):
+        client.delete_user(["u1"])
+
+    assert time.monotonic() - started < 10

tests/test_litellm/proxy/guardrails/guardrail_hooks/test_hiddenlayer.py

@@ -1,4 +1,6 @@
 import os
+import threading
+import time
 import uuid
 from typing import List, cast
 from unittest.mock import AsyncMock, MagicMock, patch
@@ -6,6 +8,7 @@ from unittest.mock import AsyncMock, MagicMock, patch
 import pytest
 from fastapi import HTTPException
 from httpx import Request, Response
+import requests
 
 
 import litellm
@@ -14,6 +17,7 @@ from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLogging
 from litellm.proxy.guardrails.guardrail_hooks.hiddenlayer.hiddenlayer import (
     HiddenlayerGuardrail,
     HiddenlayerGuardrailV2,
+    _get_jwt,
 )
 from litellm.proxy.guardrails.init_guardrails import init_guardrails_v2
 from litellm.types.utils import (
@@ -1088,3 +1092,47 @@ class TestHiddenlayerGuardrailV2:
         config_model = HiddenlayerGuardrailV2.get_config_model()
         assert config_model is not None
         assert config_model.__name__ == "HiddenlayerGuardrailConfigModel"
+
+
+@pytest.fixture
+def hanging_auth_server():
+    """A server that accepts the connection and never answers, so only a timeout ends the call."""
+    from http.server import BaseHTTPRequestHandler, HTTPServer
+    from socketserver import ThreadingMixIn
+
+    stop: threading.Event = threading.Event()
+
+    class SilentRequestHandler(BaseHTTPRequestHandler):
+        protocol_version = "HTTP/1.1"
+
+        def do_POST(self):
+            stop.wait(timeout=30)
+
+        def log_message(self, format, *args):
+            pass
+
+    class ThreadedServer(ThreadingMixIn, HTTPServer):
+        daemon_threads = True
+
+    server = ThreadedServer(("127.0.0.1", 0), SilentRequestHandler)
+    thread = threading.Thread(target=server.serve_forever, daemon=True)
+    thread.start()
+    try:
+        yield f"http://127.0.0.1:{server.server_port}"
+    finally:
+        stop.set()
+        server.shutdown()
+        server.server_close()
+        thread.join(timeout=5)
+
+
+def test_get_jwt_gives_up_at_the_timeout_instead_of_blocking_the_event_loop(hanging_auth_server):
+    """
+    `_get_jwt` runs synchronously inside `_call_hiddenlayer`, so an auth host that
+    accepts and never answers used to park the whole worker's event loop.
+    """
+    started = time.monotonic()
+    with pytest.raises(requests.exceptions.Timeout):
+        _get_jwt(auth_url=hanging_auth_server, api_id="id", api_key="secret", timeout=1)
+
+    assert time.monotonic() - started < 10

type-discipline-budget.json

@@ -1,6 +1,6 @@
 {
   "LIT001": {
-    "limit": 22805
+    "limit": 22804
   },
   "LIT002": {
     "limit": 26873
@@ -33,6 +33,6 @@
     "limit": 5588
   },
   "LIT012": {
-    "limit": 4510
+    "limit": 4508
   }
 }