[Bug]: No timeout on model download — requests.get() can hang indefinitely #627

Summary

The download_file_from_gcs method calls requests.get() with no timeout parameter. The Python requests library has no default timeout, so a stalled connection will block indefinitely. This is directly in the model initialization path, meaning any application calling TextEmbedding(model_name) or similar can hang forever.

Affected Code

File: fastembed/common/model_management.py, line 102

@classmethod
def download_file_from_gcs(cls, url: str, output_path: str, show_progress: bool = True) -> str:
    if os.path.exists(output_path):
        return output_path
    response = requests.get(url, stream=True)   # No timeout

Reproduction

# Simulate stalled server
sudo iptables -A OUTPUT -d storage.googleapis.com -j DROP

# Then in Python:
from fastembed import TextEmbedding
model = TextEmbedding("BAAI/bge-small-en-v1.5")  # Hangs indefinitely

Impact

• Application deadlock: any service using fastembed in its initialization path hangs forever if the download stalls
• No error recovery: no exception is raised, so retry logic and circuit breakers are bypassed
• Silent resource exhaustion: threads/coroutines waiting on the download are blocked indefinitely, eventually exhausting worker pools
• Especially problematic in serverless (Lambda, Cloud Run) or container environments with strict execution time limits — the process hangs until the platform kills it

Suggested Fix

Add a connect+read timeout:

response = requests.get(
    url,
    stream=True,
    timeout=(10, 60),  # (connect_timeout, read_timeout) in seconds
)

The tuple form (connect_timeout, read_timeout) is recommended: short connect timeout (10s) to detect unreachable servers, longer read timeout (60s) to allow slow transfers without hanging forever.

Found via automated codebase analysis. Confirmed independently by two reviewers (Claude, Gemini).