diff --git a/litellm/proxy/image_endpoints/endpoints.py b/litellm/proxy/image_endpoints/endpoints.py index 28fb9baf2aa..998b5b23f7f 100644 --- a/litellm/proxy/image_endpoints/endpoints.py +++ b/litellm/proxy/image_endpoints/endpoints.py @@ -6,7 +6,16 @@ from types import MappingProxyType from typing import Final, get_type_hints import orjson -from fastapi import APIRouter, Depends, File, HTTPException, Request, Response, UploadFile, status +from fastapi import ( + APIRouter, + Depends, + File, + HTTPException, + Request, + Response, + UploadFile, + status, +) from fastapi.responses import ORJSONResponse import litellm @@ -37,8 +46,12 @@ from litellm.types.llms.openai import ChatCompletionUserMessage router: Final = APIRouter() -IMAGE_EDIT_NUMERIC_FORM_FIELDS: Final = numeric_form_fields(get_type_hints(ImageEditRequestParams)) -IMAGE_EDIT_OPTIONAL_FIELD_DEFAULTS: Final = MappingProxyType({"prompt": None, "image": None}) +IMAGE_EDIT_NUMERIC_FORM_FIELDS: Final = numeric_form_fields( + get_type_hints(ImageEditRequestParams) +) +IMAGE_EDIT_OPTIONAL_FIELD_DEFAULTS: Final = MappingProxyType( + {"prompt": None, "image": None} +) IMAGE_ARRAY_FIELD: Final = "image[]" MASK_ARRAY_FIELD: Final = "mask[]" @@ -102,7 +115,9 @@ async def image_generation( version, ) - litellm_call_id: Final = resolve_litellm_call_id(request.headers.get("x-litellm-call-id")) + litellm_call_id: Final = resolve_litellm_call_id( + request.headers.get("x-litellm-call-id") + ) data = {"litellm_call_id": litellm_call_id} try: # Use orjson to parse JSON data, orjson speeds up requests significantly @@ -123,7 +138,11 @@ async def image_generation( reject_url_valued_destination("model", model) data["model"] = resolve_inference_model( - data.get("model"), general_settings, user_model, model, kind="image_generation" + data.get("model"), + general_settings, + user_model, + model, + kind="image_generation", ) ### MODEL ALIAS MAPPING ### @@ -150,6 +169,31 @@ async def image_generation( data["prompt"] = get_str_from_messages(messages) data.pop("messages", None) + ## APPLY KEY/TEAM ROUTER SETTINGS ## + # Read key/team router_settings and apply them to route_request + # This matches the behavior in base_process_llm_request + if proxy_config is not None: + try: + from litellm.proxy.common_request_processing import ( + ProxyBaseLLMRequestProcessing, + ) + + # Create a temporary processor to get router_settings_override + temp_processor = ProxyBaseLLMRequestProcessing(data=data) + router_settings_override = ( + temp_processor.common_processing_pre_call_logic( + user_api_key_dict=user_api_key_dict, + data=data, + route_type="aimage_generation", + proxy_config=proxy_config, + ) + ) + if router_settings_override: + data["router_settings_override"] = router_settings_override + except Exception: + # If reading router_settings fails, continue without override + pass + ## ROUTE TO CORRECT ENDPOINT ## llm_call: Final = await route_request( data=data, @@ -160,7 +204,11 @@ async def image_generation( response = await llm_call ### ALERTING ### - asyncio.create_task(proxy_logging_obj.update_request_status(litellm_call_id=litellm_call_id, status="success")) + asyncio.create_task( + proxy_logging_obj.update_request_status( + litellm_call_id=litellm_call_id, status="success" + ) + ) ### CALL HOOKS ### - modify outgoing data (guardrails, otel, etc.) response = await proxy_logging_obj.post_call_success_hook( @@ -191,11 +239,13 @@ async def image_generation( ) # Call response headers hook (matches base_process_llm_request behavior) - callback_headers: Final = await proxy_logging_obj.post_call_response_headers_hook( - data=data, - user_api_key_dict=user_api_key_dict, - response=response, - request_headers=dict(request.headers), + callback_headers: Final = ( + await proxy_logging_obj.post_call_response_headers_hook( + data=data, + user_api_key_dict=user_api_key_dict, + response=response, + request_headers=dict(request.headers), + ) ) if callback_headers: fastapi_response.headers.update(callback_headers) @@ -209,7 +259,9 @@ async def image_generation( if isinstance(e, HTTPException): raise ProxyException( message=getattr(e, "message", str(e)), - type=openai_error_type(e, error_status_code(e, status.HTTP_400_BAD_REQUEST)), + type=openai_error_type( + e, error_status_code(e, status.HTTP_400_BAD_REQUEST) + ), param=openai_error_param(e), headers=litellm_call_id_headers(litellm_call_id), code=error_status_code(e, status.HTTP_400_BAD_REQUEST), @@ -266,9 +318,13 @@ async def image_edit_api( ``` """ if image is not None and image_array is not None: - raise HTTPException(status_code=422, detail="Cannot specify both 'image' and 'image[]'") + raise HTTPException( + status_code=422, detail="Cannot specify both 'image' and 'image[]'" + ) if mask is not None and mask_array is not None: - raise HTTPException(status_code=422, detail="Cannot specify both 'mask' and 'mask[]'") + raise HTTPException( + status_code=422, detail="Cannot specify both 'mask' and 'mask[]'" + ) if image is None and image_array is not None: image = image_array if mask is None and mask_array is not None: @@ -303,7 +359,9 @@ async def image_edit_api( ) data: Final = { key: value - for key, value in chain(IMAGE_EDIT_OPTIONAL_FIELD_DEFAULTS.items(), form_fields.items()) + for key, value in chain( + IMAGE_EDIT_OPTIONAL_FIELD_DEFAULTS.items(), form_fields.items() + ) if key not in BRACKETED_FILE_FIELDS } image_files: Final = await batch_to_bytesio(image)