fix(proxy): clear litellm_call_id_from_client flag when no header supplied

Defensive guard: when the x-litellm-call-id header is absent, explicitly
set litellm_call_id_from_client=False on the request data dict. Without
this reset, a body-injected `"litellm_call_id_from_client": true` field
would survive into kwargs and flip the precedence in get_spend_logs_id,
causing the auto-generated UUID to be recorded as request_id instead of
the provider response.id and breaking lookups by response ID.

Applied to all three handler locations:
- common_request_processing.py (OpenAI-compatible /chat/completions, etc.)
- google_endpoints/endpoints.py:generateContent
- google_endpoints/endpoints.py:streamGenerateContent

Addresses Greptile P2 review feedback on PR #25952.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

litellm/proxy/common_request_processing.py

@@ -789,6 +789,10 @@ class ProxyBaseLLMRequestProcessing:
             self.data["litellm_call_id_from_client"] = True
         else:
             self.data["litellm_call_id"] = str(uuid.uuid4())
+            # Defensive: explicitly clear the flag so a body-injected
+            # `litellm_call_id_from_client: true` cannot flip precedence
+            # in get_spend_logs_id when no header was actually supplied.
+            self.data["litellm_call_id_from_client"] = False
         DDSpanTagger.tag_call_id(self.data.get("litellm_call_id"))
         DDSpanTagger.tag_request(
             user_api_key_dict=user_api_key_dict,

litellm/proxy/google_endpoints/endpoints.py

@@ -64,6 +64,10 @@ async def google_generate_content(
         data["litellm_call_id_from_client"] = True
     else:
         data["litellm_call_id"] = str(uuid.uuid4())
+        # Defensive: explicitly clear the flag so a body-injected
+        # `litellm_call_id_from_client: true` cannot flip precedence
+        # in get_spend_logs_id when no header was actually supplied.
+        data["litellm_call_id_from_client"] = False
     logging_obj, data = litellm.utils.function_setup(
         original_function="agenerate_content",
         rules_obj=litellm.utils.Rules(),
@@ -130,6 +134,10 @@ async def google_stream_generate_content(
         data["litellm_call_id_from_client"] = True
     else:
         data["litellm_call_id"] = str(uuid.uuid4())
+        # Defensive: explicitly clear the flag so a body-injected
+        # `litellm_call_id_from_client: true` cannot flip precedence
+        # in get_spend_logs_id when no header was actually supplied.
+        data["litellm_call_id_from_client"] = False
     logging_obj, data = litellm.utils.function_setup(
         original_function="agenerate_content_stream",
         rules_obj=litellm.utils.Rules(),