fix(types): define the responses compact call types for during_call guardrails and deployment hooks (#50274)

* fix(types): define the responses compact call types so during_call guardrails and deployment hooks see them

* test: regenerate dashboard api types and skip compact in the azure sdk client sweep

---------

Co-authored-by: devin-ai-integration[bot] <158243242+devin-ai-integration[bot]@users.noreply.github.com>

litellm/types/utils.py

@@ -703,6 +703,8 @@ class CallTypes(str, Enum):
     adelete_responses = "adelete_responses"
     cancel_responses = "cancel_responses"
     acancel_responses = "acancel_responses"
+    compact_responses = "compact_responses"
+    acompact_responses = "acompact_responses"
     alist_input_items = "alist_input_items"
     llm_passthrough_route = "llm_passthrough_route"
     allm_passthrough_route = "allm_passthrough_route"
@@ -806,6 +808,8 @@ CallTypesLiteral = Literal[
     "send_message",
     "aresponses",
     "responses",
+    "compact_responses",
+    "acompact_responses",
     "acreate_skill",
     "acreate_realtime_client_secret",
     "arealtime_calls",

tests/integration/authorization/test_responses_lifecycle_during_call_guardrail.py

@@ -0,0 +1,114 @@
+"""GET, compact, DELETE and cancel on /v1/responses answer 200 when a during_call guardrail is on by default
+(LIT-9770): the hook used to raise `'acompact_responses' is not a valid CallTypes` and the proxy returned 500"""
+
+import json
+import uuid
+from collections.abc import Iterator
+from pathlib import Path
+from typing import Final
+
+import httpx
+import pytest
+import yaml
+from integration._support.client import Gateway, string_value
+from integration._support.process import owned_proxy
+from integration._support.provider import SharedProvider
+from integration._support.wire import Reply
+from pydantic import JsonValue, TypeAdapter
+
+pytestmark: Final = pytest.mark.shares_worker_state
+
+_JSON_OBJECT: Final = TypeAdapter(dict[str, JsonValue])
+_MODEL: Final = "openai/gpt-5.4-mini"
+
+
+def _input() -> list[JsonValue]:
+    return [
+        {"role": "user", "content": f"Remember the number {uuid.uuid4().hex}."},
+        {"role": "assistant", "content": "Got it."},
+    ]
+
+
+def _response(response_id: str) -> Reply:
+    body: Final[dict[str, JsonValue]] = {
+        "id": response_id,
+        "object": "response",
+        "created_at": 1,
+        "status": "completed",
+        "model": _MODEL.removeprefix("openai/"),
+        "output": [
+            {
+                "type": "message",
+                "id": f"msg_{uuid.uuid4().hex}",
+                "status": "completed",
+                "role": "assistant",
+                "content": [{"type": "output_text", "text": "Got it.", "annotations": []}],
+            }
+        ],
+        "parallel_tool_calls": False,
+        "tool_choice": "auto",
+        "tools": [],
+        "usage": {"input_tokens": 12, "output_tokens": 5, "total_tokens": 17},
+    }
+    return Reply(body=json.dumps(body).encode())
+
+
+def _deleted(response_id: str) -> Reply:
+    return Reply(body=json.dumps({"id": response_id, "object": "response.deleted", "deleted": True}).encode())
+
+
+def _guarded_config(path: Path) -> Path:
+    base: Final = _JSON_OBJECT.validate_python(yaml.safe_load(Path("tests/integration/proxy_config.yaml").read_text()))
+    guardrail: Final[dict[str, JsonValue]] = {
+        "guardrail_name": f"content-filter-{uuid.uuid4().hex}",
+        "litellm_params": {
+            "guardrail": "litellm_content_filter",
+            "mode": "during_call",
+            "default_on": True,
+            "blocked_words": [{"keyword": "forbiddenword", "action": "BLOCK"}],
+        },
+    }
+    path.write_text(yaml.safe_dump({**base, "guardrails": [guardrail]}))
+    return path
+
+
+@pytest.fixture(scope="module")
+def guarded(session_gateway: Gateway, tmp_path_factory: pytest.TempPathFactory) -> Iterator[Gateway]:
+    directory: Final = tmp_path_factory.mktemp("during-call-guardrail")
+    with owned_proxy(session_gateway, directory, {}, config=_guarded_config(directory / "config.yaml")) as candidate:
+        yield candidate
+
+
+def _create(guarded: Gateway, provider: SharedProvider, upstream_id: str) -> str:
+    provider.expect(_response(upstream_id))
+    created: Final = guarded.request("POST", "/v1/responses", {"model": _MODEL, "input": _input(), "store": True})
+    assert created.status_code == 200, created.text
+    (upstream,) = provider.received()
+    assert upstream.target.endswith("/responses"), upstream.target
+    return string_value(_JSON_OBJECT.validate_json(created.content)["id"])
+
+
+def _lifecycle(guarded: Gateway, provider: SharedProvider, operation: str) -> tuple[httpx.Response, str]:
+    if operation == "compact":
+        provider.expect(_response(f"resp_{uuid.uuid4().hex}"))
+        return guarded.request(
+            "POST", "/v1/responses/compact", {"model": _MODEL, "input": _input()}
+        ), "/responses/compact"
+    upstream_id: Final = f"resp_{uuid.uuid4().hex}"
+    managed_id: Final = _create(guarded, provider, upstream_id)
+    provider.expect(_deleted(upstream_id) if operation == "delete" else _response(upstream_id))
+    if operation == "get":
+        return guarded.request("GET", f"/v1/responses/{managed_id}"), f"/responses/{upstream_id}"
+    if operation == "delete":
+        return guarded.request("DELETE", f"/v1/responses/{managed_id}"), f"/responses/{upstream_id}"
+    return guarded.request("POST", f"/v1/responses/{managed_id}/cancel"), f"/responses/{upstream_id}/cancel"
+
+
+@pytest.mark.parametrize("operation", ["get", "compact", "delete", "cancel"])
+def test_responses_lifecycle_answers_200_with_a_during_call_guardrail_on(
+    guarded: Gateway, provider: SharedProvider, operation: str
+) -> None:
+    response, upstream_target = _lifecycle(guarded, provider, operation)
+    assert response.status_code == 200, response.text
+    (upstream,) = provider.received()
+    assert upstream.target.endswith(upstream_target), upstream.target

tests/unit/llms/azure/test_azure_common_utils.py

@@ -477,6 +477,7 @@ def test_default_max_retries_env_var_reaches_azure_sdk_client():
             "aget_responses",
             "adelete_responses",
             "acancel_responses",
+            "acompact_responses",
             "aresponses_websocket",
             "alist_input_items",
             "acreate_fine_tuning_job",

tests/unit/proxy/guardrails/guardrail_hooks/unified_guardrails/test_unified_guardrail.py

@@ -3291,3 +3291,21 @@ async def test_pass_through_guardrail_text_is_the_client_body_without_the_inject
     texts: Final = [call["inputs"]["texts"] for call in guardrail.apply_calls]
     assert [[json.loads(text) for text in call_texts] for call_texts in texts] == [[body]], texts
     assert data["litellm_metadata"]["user_api_key_user_email"] == "someone@example.com", data["litellm_metadata"]
+
+
+@pytest.mark.asyncio
+async def test_moderation_hook_lets_responses_compact_through_without_a_translation() -> None:
+    handler: Final = UnifiedLLMGuardrails()
+    guardrail: Final = RecordingGuardrail()
+    data: Final = {
+        "guardrail_to_apply": guardrail,
+        "model": "gpt-5.4-mini",
+        "input": [{"role": "user", "content": "Remember the number 42."}],
+    }
+
+    result: Final = await handler.async_moderation_hook(
+        data=data, user_api_key_dict=UserAPIKeyAuth(api_key="test-key"), call_type="acompact_responses"
+    )
+
+    assert result is data
+    assert guardrail.apply_calls == []

tests/unit/test_utils.py

@@ -104,6 +104,7 @@ from litellm.utils import (
     _snapshot_exception_for_hook,
     async_post_call_failure_deployment_hook,
     async_post_call_success_deployment_hook,
+    async_pre_call_deployment_hook,
     calculate_max_parallel_requests,
     client,
     get_non_default_completion_params,
@@ -11626,3 +11627,25 @@ async def test_success_enqueue_is_tracked_without_starting_callbacks_synchronous
     assert GLOBAL_LOGGING_WORKER.pending_count > 0, "Tracked enqueue must finish before worker draining starts"
     await GLOBAL_LOGGING_WORKER.flush()
     assert logging_obj.model_call_details["has_logged_async_success"] is True
+
+
+class _RecordingPreCallDeploymentHook(CustomLogger):
+    def __init__(self) -> None:
+        super().__init__()
+        self.seen_call_types: tuple[CallTypes | None, ...] = ()
+
+    async def async_pre_call_deployment_hook(
+        self, kwargs: dict[str, object], call_type: CallTypes | None
+    ) -> dict[str, object] | None:
+        self.seen_call_types = (*self.seen_call_types, call_type)
+        return None
+
+
+@pytest.mark.asyncio
+async def test_pre_call_deployment_hook_hands_responses_compact_its_call_type(monkeypatch: pytest.MonkeyPatch) -> None:
+    hook: Final = _RecordingPreCallDeploymentHook()
+    monkeypatch.setattr(litellm, "callbacks", [hook])
+
+    await async_pre_call_deployment_hook({"model": "gpt-5.4-mini"}, "acompact_responses")
+
+    assert hook.seen_call_types == (CallTypes.acompact_responses,)

tests/unit/types/test_utils.py

@@ -1,7 +1,12 @@
 import pytest
 
 from litellm.litellm_core_utils.initialize_dynamic_callback_params import validate_excluded_services_value
-from litellm.types.utils import CAPTURE_MESSAGE_CONTENT_VALUES, CaptureMessageContent, captures_span_content
+from litellm.types.utils import (
+    CAPTURE_MESSAGE_CONTENT_VALUES,
+    CallTypes,
+    CaptureMessageContent,
+    captures_span_content,
+)
 
 
 @pytest.mark.parametrize("value", ["", "  ", "auth", "none,redis", "redis,", ",postgres"])
@@ -45,3 +50,8 @@ def test_capture_message_content_constructor_accepts_any_string() -> None:
 )
 def test_captures_span_content(mode: str | None, captures: bool) -> None:
     assert captures_span_content(mode) is captures
+
+
+@pytest.mark.parametrize("call_type", ["compact_responses", "acompact_responses"])
+def test_responses_compact_call_types_round_trip_their_public_names(call_type: str) -> None:
+    assert CallTypes(call_type).value == call_type

ui/litellm-dashboard/src/lib/http/schema.d.ts

@@ -29442,7 +29442,7 @@ export interface components {
          * CallTypes
          * @enum {string}
          */
-        CallTypes: "embedding" | "aembedding" | "completion" | "acompletion" | "atext_completion" | "text_completion" | "image_generation" | "aimage_generation" | "image_edit" | "aimage_edit" | "moderation" | "amoderation" | "atranscription" | "transcription" | "aspeech" | "speech" | "rerank" | "arerank" | "search" | "asearch" | "decisions" | "adecisions" | "systemone" | "asystemone" | "extraction" | "aextraction" | "_arealtime" | "_aresponses_websocket" | "acreate_realtime_client_secret" | "arealtime_calls" | "acreate_realtime_transcription_session" | "acreate_realtime_translation_client_secret" | "arealtime_translation_calls" | "create_batch" | "acreate_batch" | "aretrieve_batch" | "retrieve_batch" | "acancel_batch" | "cancel_batch" | "pass_through_endpoint" | "anthropic_messages" | "aanthropic_messages" | "get_assistants" | "aget_assistants" | "create_assistants" | "acreate_assistants" | "delete_assistant" | "adelete_assistant" | "acreate_thread" | "create_thread" | "aget_thread" | "get_thread" | "a_add_message" | "add_message" | "aget_messages" | "get_messages" | "arun_thread" | "run_thread" | "arun_thread_stream" | "run_thread_stream" | "afile_retrieve" | "file_retrieve" | "afile_delete" | "file_delete" | "afile_list" | "file_list" | "acreate_file" | "create_file" | "afile_content" | "file_content" | "create_fine_tuning_job" | "acreate_fine_tuning_job" | "create_video" | "acreate_video" | "video_generation" | "avideo_generation" | "avideo_retrieve" | "video_retrieve" | "avideo_content" | "video_content" | "video_remix" | "avideo_remix" | "video_list" | "avideo_list" | "video_retrieve_job" | "avideo_retrieve_job" | "video_delete" | "avideo_delete" | "video_create_character" | "avideo_create_character" | "video_get_character" | "avideo_get_character" | "video_edit" | "avideo_edit" | "video_extension" | "avideo_extension" | "vector_store_file_create" | "avector_store_file_create" | "vector_store_file_list" | "avector_store_file_list" | "vector_store_file_retrieve" | "avector_store_file_retrieve" | "vector_store_file_content" | "avector_store_file_content" | "vector_store_file_update" | "avector_store_file_update" | "vector_store_file_delete" | "avector_store_file_delete" | "vector_store_create" | "avector_store_create" | "vector_store_search" | "avector_store_search" | "ingest" | "aingest" | "query" | "aquery" | "create_interaction" | "acreate_interaction" | "create_container" | "acreate_container" | "list_containers" | "alist_containers" | "retrieve_container" | "aretrieve_container" | "delete_container" | "adelete_container" | "list_container_files" | "alist_container_files" | "upload_container_file" | "aupload_container_file" | "create_sandbox" | "acreate_sandbox" | "delete_sandbox" | "adelete_sandbox" | "run_code" | "arun_code" | "code_interpreter_tool" | "acode_interpreter_tool" | "acancel_fine_tuning_job" | "cancel_fine_tuning_job" | "alist_fine_tuning_jobs" | "list_fine_tuning_jobs" | "aretrieve_fine_tuning_job" | "retrieve_fine_tuning_job" | "responses" | "aresponses" | "get_responses" | "aget_responses" | "delete_responses" | "adelete_responses" | "cancel_responses" | "acancel_responses" | "alist_input_items" | "llm_passthrough_route" | "allm_passthrough_route" | "generate_content" | "agenerate_content" | "generate_content_stream" | "agenerate_content_stream" | "ocr" | "aocr" | "call_mcp_tool" | "list_mcp_tools" | "asend_message" | "send_message" | "acreate_skill";
+        CallTypes: "embedding" | "aembedding" | "completion" | "acompletion" | "atext_completion" | "text_completion" | "image_generation" | "aimage_generation" | "image_edit" | "aimage_edit" | "moderation" | "amoderation" | "atranscription" | "transcription" | "aspeech" | "speech" | "rerank" | "arerank" | "search" | "asearch" | "decisions" | "adecisions" | "systemone" | "asystemone" | "extraction" | "aextraction" | "_arealtime" | "_aresponses_websocket" | "acreate_realtime_client_secret" | "arealtime_calls" | "acreate_realtime_transcription_session" | "acreate_realtime_translation_client_secret" | "arealtime_translation_calls" | "create_batch" | "acreate_batch" | "aretrieve_batch" | "retrieve_batch" | "acancel_batch" | "cancel_batch" | "pass_through_endpoint" | "anthropic_messages" | "aanthropic_messages" | "get_assistants" | "aget_assistants" | "create_assistants" | "acreate_assistants" | "delete_assistant" | "adelete_assistant" | "acreate_thread" | "create_thread" | "aget_thread" | "get_thread" | "a_add_message" | "add_message" | "aget_messages" | "get_messages" | "arun_thread" | "run_thread" | "arun_thread_stream" | "run_thread_stream" | "afile_retrieve" | "file_retrieve" | "afile_delete" | "file_delete" | "afile_list" | "file_list" | "acreate_file" | "create_file" | "afile_content" | "file_content" | "create_fine_tuning_job" | "acreate_fine_tuning_job" | "create_video" | "acreate_video" | "video_generation" | "avideo_generation" | "avideo_retrieve" | "video_retrieve" | "avideo_content" | "video_content" | "video_remix" | "avideo_remix" | "video_list" | "avideo_list" | "video_retrieve_job" | "avideo_retrieve_job" | "video_delete" | "avideo_delete" | "video_create_character" | "avideo_create_character" | "video_get_character" | "avideo_get_character" | "video_edit" | "avideo_edit" | "video_extension" | "avideo_extension" | "vector_store_file_create" | "avector_store_file_create" | "vector_store_file_list" | "avector_store_file_list" | "vector_store_file_retrieve" | "avector_store_file_retrieve" | "vector_store_file_content" | "avector_store_file_content" | "vector_store_file_update" | "avector_store_file_update" | "vector_store_file_delete" | "avector_store_file_delete" | "vector_store_create" | "avector_store_create" | "vector_store_search" | "avector_store_search" | "ingest" | "aingest" | "query" | "aquery" | "create_interaction" | "acreate_interaction" | "create_container" | "acreate_container" | "list_containers" | "alist_containers" | "retrieve_container" | "aretrieve_container" | "delete_container" | "adelete_container" | "list_container_files" | "alist_container_files" | "upload_container_file" | "aupload_container_file" | "create_sandbox" | "acreate_sandbox" | "delete_sandbox" | "adelete_sandbox" | "run_code" | "arun_code" | "code_interpreter_tool" | "acode_interpreter_tool" | "acancel_fine_tuning_job" | "cancel_fine_tuning_job" | "alist_fine_tuning_jobs" | "list_fine_tuning_jobs" | "aretrieve_fine_tuning_job" | "retrieve_fine_tuning_job" | "responses" | "aresponses" | "get_responses" | "aget_responses" | "delete_responses" | "adelete_responses" | "cancel_responses" | "acancel_responses" | "compact_responses" | "acompact_responses" | "alist_input_items" | "llm_passthrough_route" | "allm_passthrough_route" | "generate_content" | "agenerate_content" | "generate_content_stream" | "agenerate_content_stream" | "ocr" | "aocr" | "call_mcp_tool" | "list_mcp_tools" | "asend_message" | "send_message" | "acreate_skill";
         /** CallbackDelete */
         CallbackDelete: {
             /** Callback Name */