diff --git a/backend/app/services/gemini_service.py b/backend/app/services/gemini_service.py
index 4401541..d15c659 100755
--- a/backend/app/services/gemini_service.py
+++ b/backend/app/services/gemini_service.py
@@ -11,11 +11,10 @@ from app.models.schemas import SubReview, RagStatus
 # Configure logging
 logger = logging.getLogger(__name__)
 
-# Timeout (seconds) for each Gemini API call.
-# Set at the HTTP client level so the network connection is actually closed
-# on timeout — asyncio.wait_for alone cannot cancel SDK-internal retries.
-_PRIMARY_TIMEOUT = 45
-_FALLBACK_TIMEOUT = 150
+# Timeout for each Gemini API call.
+# google-genai http_options 'timeout' is in MILLISECONDS.
+_PRIMARY_TIMEOUT_MS = 45_000   # 45 seconds
+_FALLBACK_TIMEOUT_MS = 150_000  # 150 seconds
 
 
 class GeminiService:
@@ -30,13 +29,14 @@ class GeminiService:
         """
         # Two separate clients with different HTTP-level timeouts so the
         # network connection is torn down when the deadline is reached.
+        # Note: google-genai http_options 'timeout' is in milliseconds.
         self.primary_client = genai.Client(
             api_key=api_key,
-            http_options={"timeout": _PRIMARY_TIMEOUT},
+            http_options={"timeout": _PRIMARY_TIMEOUT_MS},
         )
         self.fallback_client = genai.Client(
             api_key=api_key,
-            http_options={"timeout": _FALLBACK_TIMEOUT},
+            http_options={"timeout": _FALLBACK_TIMEOUT_MS},
         )
         self.model = "gemini-3.1-pro-preview"
         self.fallback_model = "gemini-3-flash-preview"