From db1b1e219551bc53bec9ed6fb60cb90502dd87ce Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 09:45:52 +0000
Subject: [PATCH] fix: bound Hugging Face config fetch and keep embedding tests
 off the network

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 litellm/constants.py                                |  1 +
 litellm/utils.py                                    |  5 +++--
 .../embedding/test_huggingface_embedding_handler.py | 13 ++++++++++++-
 3 files changed, 16 insertions(+), 3 deletions(-)

diff --git a/litellm/constants.py b/litellm/constants.py
index fc88086805f..aa7ab1d81ff 100644
--- a/litellm/constants.py
+++ b/litellm/constants.py
@@ -35,6 +35,7 @@ DEFAULT_COOLDOWN_TIME_SECONDS: Final = int(os.getenv("DEFAULT_COOLDOWN_TIME_SECO
 DEFAULT_REPLICATE_POLLING_RETRIES: Final = int(os.getenv("DEFAULT_REPLICATE_POLLING_RETRIES", 5))
 DEFAULT_REPLICATE_POLLING_DELAY_SECONDS: Final = int(os.getenv("DEFAULT_REPLICATE_POLLING_DELAY_SECONDS", 1))
 DEFAULT_IMAGE_TOKEN_COUNT: Final = int(os.getenv("DEFAULT_IMAGE_TOKEN_COUNT", 250))
+HF_CONFIG_FETCH_TIMEOUT_SECONDS: Final = float(os.getenv("HF_CONFIG_FETCH_TIMEOUT_SECONDS", 10))
 
 # Maximum wall-clock seconds a streaming response is allowed to run.
 # Streams exceeding this duration are terminated with a Timeout error.
diff --git a/litellm/utils.py b/litellm/utils.py
index 5cd9bfc5f32..1f4674fedf0 100644
--- a/litellm/utils.py
+++ b/litellm/utils.py
@@ -69,6 +69,7 @@ from litellm.constants import (
     DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
     DEFAULT_TRIM_RATIO,
     FUNCTION_DEFINITION_TOKEN_COUNT,
+    HF_CONFIG_FETCH_TIMEOUT_SECONDS,
     INITIAL_RETRY_DELAY,
     JITTER,
     MAX_RETRY_DELAY,
@@ -5168,7 +5169,7 @@ def get_max_tokens(model: str) -> int | None:
         config_url: Final = f"https://huggingface.co/{model_name}/raw/main/config.json"
         try:
             # Make the HTTP request to get the raw JSON file
-            response: Final = litellm.module_level_client.get(config_url)
+            response: Final = litellm.module_level_client.get(config_url, timeout=HF_CONFIG_FETCH_TIMEOUT_SECONDS)
             response.raise_for_status()  # Raise an exception for bad responses (4xx or 5xx)
 
             # Parse the JSON response
@@ -5522,7 +5523,7 @@ def _get_max_position_embeddings(model_name: str) -> int | None:
 
     try:
         # Make the HTTP request to get the raw JSON file
-        response: Final = litellm.module_level_client.get(config_url)
+        response: Final = litellm.module_level_client.get(config_url, timeout=HF_CONFIG_FETCH_TIMEOUT_SECONDS)
         response.raise_for_status()  # Raise an exception for bad responses (4xx or 5xx)
 
         # Parse the JSON response
diff --git a/tests/test_litellm/llms/huggingface/embedding/test_huggingface_embedding_handler.py b/tests/test_litellm/llms/huggingface/embedding/test_huggingface_embedding_handler.py
index f1226311b5e..0384fb796d9 100644
--- a/tests/test_litellm/llms/huggingface/embedding/test_huggingface_embedding_handler.py
+++ b/tests/test_litellm/llms/huggingface/embedding/test_huggingface_embedding_handler.py
@@ -4,6 +4,7 @@ from unittest.mock import patch, MagicMock, AsyncMock
 
 import litellm
 import pytest
+import respx
 
 
 MOCK_EMBEDDING_RESPONSE = [[0.1, 0.2, 0.3, 0.4, 0.5]]
@@ -21,6 +22,16 @@ def mock_embedding_http_handler():
         yield mock_post
 
 
+@pytest.fixture
+def mock_hf_config_fetch():
+    """Serve the Hugging Face config.json fetched during cost calculation, so no test leaves the process"""
+    with respx.mock(assert_all_called=False) as respx_mock:
+        respx_mock.get(url__regex=r"https://huggingface\.co/.*/config\.json").respond(
+            json={"max_position_embeddings": 512}
+        )
+        yield respx_mock
+
+
 @pytest.fixture
 def mock_embedding_async_http_handler():
     """Fixture to mock the async HTTP handler for embedding tests"""
@@ -39,7 +50,7 @@ def mock_embedding_async_http_handler():
 
 class TestHuggingFaceEmbedding:
     @pytest.fixture(autouse=True)
-    def setup(self, mock_embedding_http_handler, mock_embedding_async_http_handler):
+    def setup(self, mock_embedding_http_handler, mock_embedding_async_http_handler, mock_hf_config_fetch):
         self.mock_get_task_patcher = patch(
             "litellm.llms.huggingface.embedding.handler.get_hf_task_embedding_for_model"
         )