fix(gcs_bucket): store request payloads under the spend log request id so the logs viewer can find them

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

basedpyright-code-budget.json

@@ -1,6 +1,6 @@
 {
   "reportAny": {
-    "limit": 14076
+    "limit": 14075
   },
   "reportArgumentType": {
     "limit": 2216
@@ -57,7 +57,7 @@
     "limit": 5601
   },
   "reportMissingTypeArgument": {
-    "limit": 15306
+    "limit": 15304
   },
   "reportMissingTypeStubs": {
     "limit": 40
@@ -105,13 +105,13 @@
     "limit": 109
   },
   "reportUnknownMemberType": {
-    "limit": 38350
+    "limit": 38348
   },
   "reportUnknownParameterType": {
-    "limit": 19626
+    "limit": 19624
   },
   "reportUnknownVariableType": {
-    "limit": 29890
+    "limit": 29888
   },
   "reportUnnecessaryCast": {
     "limit": 111

litellm/integrations/gcs_bucket/gcs_bucket.py

@@ -14,6 +14,7 @@ from litellm.integrations.gcs_bucket.gcs_bucket_base import GCSBucketBase
 from litellm.litellm_core_utils.cloud_storage_security import (
     sanitize_cloud_object_component,
 )
+from litellm.litellm_core_utils.spend_log_request_id import get_spend_logs_id
 from litellm.proxy._types import CommonProxyErrors
 from litellm.types.integrations.base_health_check import IntegrationHealthCheckStatus
 from litellm.types.integrations.gcs_bucket import *
@@ -289,7 +290,7 @@ class GCSBucketLogger(GCSBucketBase, AdditionalLoggingUtils):
         else:
             object_name = self._generate_success_object_name(
                 request_date_str=current_date,
-                response_id=response_obj.get("id", ""),
+                response_id=get_spend_logs_id(kwargs.get("call_type") or "acompletion", response_obj, kwargs) or "",
             )
 
         # used for testing

litellm/litellm_core_utils/spend_log_request_id.py

@@ -0,0 +1,59 @@
+"""The id a request is filed under in LiteLLM_SpendLogs and in the per-request payload
+stores (GCS) the logs viewer reads back through ``/spend/logs/ui/{request_id}``.
+
+Both sides must derive the id the same way or the viewer looks up a payload under a key
+it was never stored under, so the derivation lives here rather than in either caller.
+"""
+
+from __future__ import annotations
+
+from collections.abc import Mapping
+from typing import Final
+
+from litellm.types.utils import CallTypes
+
+BATCH_COST_REQUEST_ID_SUFFIX: Final = "_batch_cost"
+
+_RESPONSE_ID_KEYED_CALL_TYPES: Final = frozenset(
+    {
+        CallTypes.acreate_batch.value,
+        CallTypes.aretrieve_batch.value,
+        CallTypes.acreate_file.value,
+    }
+)
+"""Batch and file rows key off the object's own id so repeated polls of the same
+object collapse into one row instead of billing it once per poll. Every other call
+type keys off the proxy-generated per-call id: request_id is the LiteLLM_SpendLogs
+primary key and the flush inserts with skip_duplicates, so keying off the provider's
+response id silently drops every row after the first whenever a provider (commonly a
+self-hosted OpenAI-compatible server) reuses completion ids."""
+
+
+def _standard_logging_id(kwargs: Mapping[str, object]) -> str | None:
+    match kwargs.get("standard_logging_object"):
+        case {"id": str() as standard_logging_id}:
+            return standard_logging_id
+        case _:
+            return None
+
+
+def get_spend_logs_id(call_type: str, response_obj: Mapping[str, object], kwargs: Mapping[str, object]) -> str | None:
+    candidate_ids: Final = (
+        (
+            response_obj.get("id"),
+            _standard_logging_id(kwargs),
+            kwargs.get("litellm_call_id"),
+        )
+        if call_type in _RESPONSE_ID_KEYED_CALL_TYPES
+        else (
+            kwargs.get("litellm_call_id"),
+            _standard_logging_id(kwargs),
+            response_obj.get("id"),
+        )
+    )
+    resolved_id: Final = next(
+        (candidate for candidate in candidate_ids if isinstance(candidate, str) and candidate), None
+    )
+    if resolved_id is not None and call_type == CallTypes.aretrieve_batch.value:
+        return f"{resolved_id}{BATCH_COST_REQUEST_ID_SUFFIX}"
+    return resolved_id

litellm/proxy/spend_tracking/spend_tracking_utils.py

@@ -30,11 +30,13 @@ from litellm.litellm_core_utils.litellm_logging import (
     request_model_access_groups_from_litellm_params,
 )
 from litellm.litellm_core_utils.safe_json_dumps import safe_dumps, strip_null_bytes
+from litellm.litellm_core_utils.spend_log_request_id import (
+    get_spend_logs_id,
+)
 from litellm.proxy._types import SpendLogsMetadata, SpendLogsPayload, SpendLogsRouterMetadata
 from litellm.proxy.spend_tracking.spend_log_error_logger import spend_log_error
 from litellm.proxy.utils import PrismaClient, hash_token
 from litellm.types.utils import (
-    CallTypes,
     CostBreakdown,
     StandardLoggingGuardrailInformation,
     StandardLoggingMCPToolCall,
@@ -211,49 +213,6 @@ def _get_spend_logs_metadata(
     return clean_metadata
 
 
-BATCH_COST_REQUEST_ID_SUFFIX: Final = "_batch_cost"
-
-_RESPONSE_ID_KEYED_CALL_TYPES: Final = frozenset(
-    {
-        CallTypes.acreate_batch.value,
-        CallTypes.aretrieve_batch.value,
-        CallTypes.acreate_file.value,
-    }
-)
-"""Batch and file rows key off the object's own id so repeated polls of the same
-object collapse into one row instead of billing it once per poll. Every other call
-type keys off the proxy-generated per-call id: request_id is the LiteLLM_SpendLogs
-primary key and the flush inserts with skip_duplicates, so keying off the provider's
-response id silently drops every row after the first whenever a provider (commonly a
-self-hosted OpenAI-compatible server) reuses completion ids."""
-
-
-def get_spend_logs_id(call_type: str, response_obj: dict, kwargs: dict) -> str | None:
-    standard_logging_payload = kwargs.get("standard_logging_object")
-    standard_logging_id: Final = (
-        standard_logging_payload.get("id") if isinstance(standard_logging_payload, dict) else None
-    )
-    candidate_ids: Final = (
-        (
-            response_obj.get("id"),
-            standard_logging_id,
-            kwargs.get("litellm_call_id"),
-        )
-        if call_type in _RESPONSE_ID_KEYED_CALL_TYPES
-        else (
-            kwargs.get("litellm_call_id"),
-            standard_logging_id,
-            response_obj.get("id"),
-        )
-    )
-    resolved_id: Final = next(
-        (candidate for candidate in candidate_ids if isinstance(candidate, str) and candidate), None
-    )
-    if resolved_id is not None and call_type == CallTypes.aretrieve_batch.value:
-        return f"{resolved_id}{BATCH_COST_REQUEST_ID_SUFFIX}"
-    return resolved_id
-
-
 _MISSING_ATTRIBUTE: Final = object()
 
 

tests/test_litellm/integrations/gcs_bucket/test_gcs_bucket_base.py

@@ -1,10 +1,13 @@
+import json
 import os
+from datetime import datetime, timezone
 from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
 
 from litellm.integrations.gcs_bucket.gcs_bucket import GCSBucketLogger
 from litellm.integrations.gcs_bucket.gcs_bucket_base import GCSBucketBase
+from litellm.litellm_core_utils.spend_log_request_id import get_spend_logs_id
 
 
 class TestGCSBucketBase:
@@ -128,3 +131,26 @@ class TestGCSBucketBase:
         assert object_name.endswith("-target_uploadType_media")
         assert ".." not in object_name
         assert "?" not in object_name
+
+    @pytest.mark.asyncio
+    async def test_logs_viewer_finds_payloads_stored_for_requests_sharing_a_provider_response_id(self):
+        logger = GCSBucketLogger.__new__(GCSBucketLogger)
+        provider_response = {"id": "chatcmpl-reused"}
+        stored_objects = {
+            logger._get_object_name(
+                kwargs={"call_type": "acompletion", "litellm_call_id": call_id},
+                logging_payload={"id": "chatcmpl-reused"},
+                response_obj=provider_response,
+            ): json.dumps({"litellm_call_id": call_id})
+            for call_id in ("call-id-1", "call-id-2")
+        }
+        assert len(stored_objects) == 2
+        logger.download_gcs_object = AsyncMock(side_effect=lambda object_name: stored_objects.get(object_name))
+
+        payload = await logger.get_request_response_payload(
+            request_id=get_spend_logs_id("acompletion", provider_response, {"litellm_call_id": "call-id-2"}),
+            start_time_utc=datetime.now(timezone.utc),
+            end_time_utc=None,
+        )
+
+        assert payload == {"litellm_call_id": "call-id-2"}

type-discipline-budget.json

@@ -1,6 +1,6 @@
 {
   "LIT001": {
-    "limit": 22367
+    "limit": 22365
   },
   "LIT002": {
     "limit": 26777
@@ -27,7 +27,7 @@
     "limit": 0
   },
   "LIT010": {
-    "limit": 16507
+    "limit": 16506
   },
   "LIT011": {
     "limit": 5535