From 3b1ebbc7cb9b6431f38b534c439f97d1cfce691c Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Fri, 17 Jul 2026 15:55:34 +0000
Subject: [PATCH] fix(responses): warn when spend-log session query hits row
 cap

---
 .../session_handler.py                        |  9 +++++
 .../test_session_handler.py                   | 37 +++++++++++++++++++
 2 files changed, 46 insertions(+)

diff --git a/litellm/responses/litellm_completion_transformation/session_handler.py b/litellm/responses/litellm_completion_transformation/session_handler.py
index af341ed9c09..e9b0450bcc5 100644
--- a/litellm/responses/litellm_completion_transformation/session_handler.py
+++ b/litellm/responses/litellm_completion_transformation/session_handler.py
@@ -295,6 +295,15 @@ class ResponsesSessionHandler:
             query, previous_response_id, DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION
         )
 
+        if len(spend_logs) >= DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION:
+            verbose_proxy_logger.warning(
+                "Responses session for previous_response_id=%s hit the %d-row cap "
+                "(DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION); the oldest turns were dropped. "
+                "Raise the limit if you need more history.",
+                previous_response_id,
+                DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION,
+            )
+
         verbose_proxy_logger.debug(
             "Found the following spend logs for previous response id %s: %s",
             previous_response_id,
diff --git a/tests/test_litellm/responses/litellm_completion_transformation/test_session_handler.py b/tests/test_litellm/responses/litellm_completion_transformation/test_session_handler.py
index 9cfc8ac2e4c..dd26dfc93e7 100644
--- a/tests/test_litellm/responses/litellm_completion_transformation/test_session_handler.py
+++ b/tests/test_litellm/responses/litellm_completion_transformation/test_session_handler.py
@@ -491,3 +491,40 @@ async def test_get_all_spend_logs_returns_most_recent_in_ascending_order():
     executed_query = mock_prisma_client.db.query_raw.await_args.args[0]
     assert 'ORDER BY "endTime" DESC' in executed_query
     assert executed_query.rstrip().rstrip(";").endswith('ORDER BY "endTime" ASC')
+
+
+@pytest.mark.asyncio
+async def test_get_all_spend_logs_warns_when_cap_is_hit():
+    """
+    Operators need a runtime signal when a session is truncated: when the query returns
+    the full cap of rows (older turns silently dropped), a warning must be emitted.
+    """
+    cap = litellm.constants.DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION
+    mock_prisma_client = AsyncMock()
+    mock_prisma_client.db.query_raw = AsyncMock(return_value=[{} for _ in range(cap)])
+
+    with _patched_prisma_client(mock_prisma_client):
+        with patch.object(session_handler.verbose_proxy_logger, "warning") as mock_warning:
+            await ResponsesSessionHandler.get_all_spend_logs_for_previous_response_id(
+                "resp_previous_id"
+            )
+
+    mock_warning.assert_called_once()
+
+
+@pytest.mark.asyncio
+async def test_get_all_spend_logs_does_not_warn_below_cap():
+    """
+    Below the cap nothing is truncated, so no truncation warning should be emitted.
+    """
+    cap = litellm.constants.DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION
+    mock_prisma_client = AsyncMock()
+    mock_prisma_client.db.query_raw = AsyncMock(return_value=[{} for _ in range(cap - 1)])
+
+    with _patched_prisma_client(mock_prisma_client):
+        with patch.object(session_handler.verbose_proxy_logger, "warning") as mock_warning:
+            await ResponsesSessionHandler.get_all_spend_logs_for_previous_response_id(
+                "resp_previous_id"
+            )
+
+    mock_warning.assert_not_called()