Pricing
-
-
+ Input: ${modelData.input_cost}/1M tokens
-Output: ${modelData.output_cost}/1M tokens
-
+ {showInput &&
+ );
+}
From 06fb9dc1caf7490e017e48378aeebb449350a8a3 Mon Sep 17 00:00:00 2001
From: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
Date: Thu, 3 Sep 2026 00:44:53 -0700
Subject: [PATCH 002/278] fix(images): stop forwarding the raw image[] and
mask[] form keys
The /v1/images/edits handler binds the documented image[] and mask[] aliases
into their canonical parameters, then re-reads the multipart body, so the raw
bracketed keys rode along to the provider next to the values already built from
them. OpenAI rejected both: image[] as "Invalid type for 'image[0]'" and mask[]
as "Invalid parameter: 'mask'". Drop both aliases from what gets forwarded.
---
litellm/proxy/image_endpoints/endpoints.py | 14 +++-
.../proxy/image_endpoints/test_endpoints.py | 75 +++++++++++++++++++
2 files changed, 86 insertions(+), 3 deletions(-)
diff --git a/litellm/proxy/image_endpoints/endpoints.py b/litellm/proxy/image_endpoints/endpoints.py
index 83caa92ede5..7d6d37c7c75 100644
--- a/litellm/proxy/image_endpoints/endpoints.py
+++ b/litellm/proxy/image_endpoints/endpoints.py
@@ -21,6 +21,10 @@ from litellm.types.llms.openai import ChatCompletionUserMessage
router: Final = APIRouter()
+IMAGE_ARRAY_FIELD: Final = "image[]"
+MASK_ARRAY_FIELD: Final = "mask[]"
+BRACKETED_FILE_FIELDS: Final = frozenset({IMAGE_ARRAY_FIELD, MASK_ARRAY_FIELD})
+
async def uploadfile_to_bytesio(upload: UploadFile) -> io.BytesIO:
"""
@@ -229,9 +233,9 @@ async def image_edit_api(
fastapi_response: Response,
user_api_key_dict: UserAPIKeyAuth = Depends(user_api_key_auth),
image: list[UploadFile] | None = File(None),
- image_array: list[UploadFile] | None = File(None, alias="image[]"),
+ image_array: list[UploadFile] | None = File(None, alias=IMAGE_ARRAY_FIELD),
mask: list[UploadFile] | None = File(None),
- mask_array: list[UploadFile] | None = File(None, alias="mask[]"),
+ mask_array: list[UploadFile] | None = File(None, alias=MASK_ARRAY_FIELD),
model: str | None = None,
):
"""
@@ -279,7 +283,11 @@ async def image_edit_api(
#########################################################
# Read request body and convert UploadFiles to BytesIO
#########################################################
- data: Final = await _read_request_body(request=request)
+ data: Final = {
+ key: value
+ for key, value in (await _read_request_body(request=request)).items()
+ if key not in BRACKETED_FILE_FIELDS
+ }
image_files: Final = await batch_to_bytesio(image)
mask_files: Final = await batch_to_bytesio(mask)
if image_files:
diff --git a/tests/test_litellm/proxy/image_endpoints/test_endpoints.py b/tests/test_litellm/proxy/image_endpoints/test_endpoints.py
index 91a011a8234..65524b54742 100644
--- a/tests/test_litellm/proxy/image_endpoints/test_endpoints.py
+++ b/tests/test_litellm/proxy/image_endpoints/test_endpoints.py
@@ -5,10 +5,13 @@ from typing import Any, Dict
import orjson
import pytest
+from fastapi import FastAPI
+from fastapi.testclient import TestClient
from starlette.requests import Request
from starlette.responses import Response
from litellm.proxy._types import UserAPIKeyAuth
+from litellm.proxy.auth.user_api_key_auth import user_api_key_auth
from litellm.proxy.image_endpoints import endpoints
@@ -115,3 +118,75 @@ async def test_image_generation_prompt_rerouting(monkeypatch):
assert captured_route_request_data["prompt"] == "sanitized prompt"
assert "messages" not in captured_route_request_data
assert response.headers.get("x-callback-test") == "value"
+
+
+def _image_edit_client(monkeypatch, captured: Dict[str, Any]) -> TestClient:
+ class CaptureProcessing:
+ def __init__(self, data: Dict[str, Any]) -> None:
+ captured.update(data)
+
+ async def base_process_llm_request(self, **_: Any) -> Dict[str, Any]:
+ return {"data": [{"b64_json": "aGk="}]}
+
+ monkeypatch.setattr(endpoints, "ProxyBaseLLMRequestProcessing", CaptureProcessing)
+ monkeypatch.setattr("litellm.proxy.proxy_server.general_settings", {})
+ monkeypatch.setattr("litellm.proxy.proxy_server.user_model", None)
+
+ app = FastAPI()
+ app.include_router(endpoints.router)
+ app.dependency_overrides[user_api_key_auth] = lambda: UserAPIKeyAuth()
+ return TestClient(app)
+
+
+def test_image_edit_image_array_alias_is_not_forwarded(monkeypatch):
+ """The documented `image[]` alias must reach the provider only as `image`."""
+ captured: Dict[str, Any] = {}
+
+ response = _image_edit_client(monkeypatch, captured).post(
+ "/v1/images/edits",
+ files={"image[]": ("tree.png", b"\x89PNG\r\n\x1a\ntree", "image/png")},
+ data={"model": "gpt-image-1", "prompt": "add a hat"},
+ )
+
+ assert response.status_code == 200
+ assert "image[]" not in captured
+ assert [buffer.getvalue() for buffer in captured["image"]] == [b"\x89PNG\r\n\x1a\ntree"]
+ assert [buffer.name for buffer in captured["image"]] == ["tree.png"]
+
+
+def test_image_edit_mask_array_alias_is_not_forwarded(monkeypatch):
+ """`mask[]` has the same shape as `image[]` and must be dropped the same way."""
+ captured: Dict[str, Any] = {}
+
+ response = _image_edit_client(monkeypatch, captured).post(
+ "/v1/images/edits",
+ files={
+ "image": ("tree.png", b"\x89PNG\r\n\x1a\ntree", "image/png"),
+ "mask[]": ("mask.png", b"\x89PNG\r\n\x1a\nmask", "image/png"),
+ },
+ data={"model": "gpt-image-1", "prompt": "add a hat"},
+ )
+
+ assert response.status_code == 200
+ assert "mask[]" not in captured
+ assert [buffer.getvalue() for buffer in captured["mask"]] == [b"\x89PNG\r\n\x1a\nmask"]
+ assert [buffer.getvalue() for buffer in captured["image"]] == [b"\x89PNG\r\n\x1a\ntree"]
+
+
+def test_image_edit_canonical_file_fields_still_reach_the_provider(monkeypatch):
+ """Dropping the bracketed aliases must not touch the canonical fields."""
+ captured: Dict[str, Any] = {}
+
+ response = _image_edit_client(monkeypatch, captured).post(
+ "/v1/images/edits",
+ files={
+ "image": ("tree.png", b"\x89PNG\r\n\x1a\ntree", "image/png"),
+ "mask": ("mask.png", b"\x89PNG\r\n\x1a\nmask", "image/png"),
+ },
+ data={"model": "gpt-image-1", "prompt": "add a hat"},
+ )
+
+ assert response.status_code == 200
+ assert [buffer.getvalue() for buffer in captured["image"]] == [b"\x89PNG\r\n\x1a\ntree"]
+ assert [buffer.getvalue() for buffer in captured["mask"]] == [b"\x89PNG\r\n\x1a\nmask"]
+ assert captured["prompt"] == "add a hat"
From 0cd14b83b33d710f0e93e9c9ee90d380ed57dfb0 Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 14:25:18 +0000
Subject: [PATCH 003/278] fix(responses): honor nested additional_drop_params
paths
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
litellm/responses/utils.py | 24 +++++---
.../responses/test_responses_utils.py | 57 +++++++++++++++++++
2 files changed, 74 insertions(+), 7 deletions(-)
diff --git a/litellm/responses/utils.py b/litellm/responses/utils.py
index 599e978df6a..677c4950f96 100644
--- a/litellm/responses/utils.py
+++ b/litellm/responses/utils.py
@@ -1,6 +1,7 @@
import base64
import re
from collections.abc import Iterable, Mapping, Sequence
+from functools import reduce
from typing import Any, Final, Optional, TypeVar, Union, cast, get_type_hints, overload
from pydantic import BaseModel
@@ -8,6 +9,7 @@ from typing_extensions import TypeIs # noqa: TID251 # narrows untyped wire pay
import litellm
from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.dot_notation_indexing import delete_nested_value, is_nested_path
from litellm.llms.base_llm.responses.transformation import BaseResponsesAPIConfig
from litellm.types.llms.openai import (
AllMessageValues,
@@ -29,6 +31,11 @@ from litellm.types.utils import (
)
+def _apply_nested_drop_params(params: dict[str, Any], additional_drop_params: list[str] | None) -> dict[str, Any]:
+ nested_paths: Final = tuple(path for path in additional_drop_params or () if is_nested_path(path))
+ return reduce(lambda acc, path: delete_nested_value(acc, path), nested_paths, params)
+
+
def _output_token_detail(details: object, field: str) -> int | None:
value: Final = getattr(details, field, None)
return value if isinstance(value, int) else None
@@ -265,13 +272,16 @@ class ResponsesAPIRequestUtils:
special_params: Final[dict[str, object]] = params.pop("kwargs", {})
additional_drop_params: Final[list[str] | None] = params.pop("additional_drop_params", None)
- non_default_params: Final = PreProcessNonDefaultParams.base_pre_process_non_default_params(
- passed_params=params,
- special_params=special_params,
- custom_llm_provider=custom_llm_provider,
- additional_drop_params=additional_drop_params,
- default_param_values={k: None for k in valid_keys},
- additional_endpoint_specific_params=["input"],
+ non_default_params: Final = _apply_nested_drop_params(
+ PreProcessNonDefaultParams.base_pre_process_non_default_params(
+ passed_params=params,
+ special_params=special_params,
+ custom_llm_provider=custom_llm_provider,
+ additional_drop_params=additional_drop_params,
+ default_param_values={k: None for k in valid_keys},
+ additional_endpoint_specific_params=["input"],
+ ),
+ additional_drop_params,
)
# decode previous_response_id if it's a litellm encoded id
diff --git a/tests/test_litellm/responses/test_responses_utils.py b/tests/test_litellm/responses/test_responses_utils.py
index 9d9eefdceb3..f1d197c1c22 100644
--- a/tests/test_litellm/responses/test_responses_utils.py
+++ b/tests/test_litellm/responses/test_responses_utils.py
@@ -119,6 +119,63 @@ class TestResponsesAPIRequestUtils:
assert result["max_output_tokens"] == 100
assert result["prompt"] == {"id": "pmpt_456"}
+ def test_get_requested_response_api_optional_param_drops_nested_path(self):
+ """Nested additional_drop_params paths like reasoning.summary must be honored"""
+ params = {
+ "temperature": 0.1,
+ "reasoning": {"effort": "high", "summary": "auto"},
+ "additional_drop_params": ["reasoning.summary"],
+ }
+
+ result = ResponsesAPIRequestUtils.get_requested_response_api_optional_param(params)
+
+ assert result["reasoning"] == {"effort": "high"}
+ assert result["temperature"] == 0.1
+
+ def test_get_requested_response_api_optional_param_drops_array_path(self):
+ """Array wildcard paths like tools[*].input_examples must be honored"""
+ params = {
+ "tools": [{"type": "function", "name": "t", "input_examples": ["x"]}],
+ "additional_drop_params": ["tools[*].input_examples"],
+ }
+
+ result = ResponsesAPIRequestUtils.get_requested_response_api_optional_param(params)
+
+ assert result["tools"] == [{"type": "function", "name": "t"}]
+
+ def test_get_requested_response_api_optional_param_drops_top_level(self):
+ """Top-level additional_drop_params keys must still be honored"""
+ params = {
+ "reasoning": {"effort": "high", "summary": "auto"},
+ "additional_drop_params": ["reasoning"],
+ }
+
+ result = ResponsesAPIRequestUtils.get_requested_response_api_optional_param(params)
+
+ assert "reasoning" not in result
+
+ def test_get_requested_response_api_optional_param_non_matching_nested_path(self):
+ """A nested path that does not match anything leaves params untouched"""
+ params = {
+ "reasoning": {"effort": "high", "summary": "auto"},
+ "additional_drop_params": ["reasoning.nope"],
+ }
+
+ result = ResponsesAPIRequestUtils.get_requested_response_api_optional_param(params)
+
+ assert result["reasoning"] == {"effort": "high", "summary": "auto"}
+
+ def test_get_requested_response_api_optional_param_none_drop_params(self):
+ """additional_drop_params=None is a no-op"""
+ params = {
+ "reasoning": {"effort": "high", "summary": "auto"},
+ "additional_drop_params": None,
+ }
+
+ result = ResponsesAPIRequestUtils.get_requested_response_api_optional_param(params)
+
+ assert result["reasoning"] == {"effort": "high", "summary": "auto"}
+
def test_decode_previous_response_id_to_original_previous_response_id(self):
"""Test decoding a LiteLLM encoded previous_response_id to the original previous_response_id"""
# Setup
From 1c22ae4805bc845fc84db76d4a53513137be18b2 Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 14:34:45 +0000
Subject: [PATCH 004/278] fix(responses): drop Any from nested drop params
helper
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
litellm/responses/utils.py | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/litellm/responses/utils.py b/litellm/responses/utils.py
index 677c4950f96..42bd99ccae2 100644
--- a/litellm/responses/utils.py
+++ b/litellm/responses/utils.py
@@ -31,7 +31,7 @@ from litellm.types.utils import (
)
-def _apply_nested_drop_params(params: dict[str, Any], additional_drop_params: list[str] | None) -> dict[str, Any]:
+def _apply_nested_drop_params(params: dict[str, object], additional_drop_params: list[str] | None) -> dict[str, object]:
nested_paths: Final = tuple(path for path in additional_drop_params or () if is_nested_path(path))
return reduce(lambda acc, path: delete_nested_value(acc, path), nested_paths, params)
@@ -288,7 +288,7 @@ class ResponsesAPIRequestUtils:
if "previous_response_id" in non_default_params:
decoded_previous_response_id: Final = (
ResponsesAPIRequestUtils.decode_previous_response_id_to_original_previous_response_id(
- non_default_params["previous_response_id"]
+ cast(str, non_default_params["previous_response_id"])
)
)
non_default_params["previous_response_id"] = decoded_previous_response_id
@@ -296,7 +296,9 @@ class ResponsesAPIRequestUtils:
if "metadata" in non_default_params:
from litellm.utils import add_openai_metadata
- converted_metadata: Final = add_openai_metadata(non_default_params["metadata"])
+ converted_metadata: Final = add_openai_metadata(
+ cast(Mapping[str, object] | None, non_default_params["metadata"])
+ )
if converted_metadata is not None:
non_default_params["metadata"] = converted_metadata
else:
From 96b7fbb1fc39a22beba061c123126f7a71ee1575 Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Fri, 11 Sep 2026 14:39:44 +0000
Subject: [PATCH 005/278] fix(responses): narrow response params without cast
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
litellm/responses/utils.py | 10 +++++-----
1 file changed, 5 insertions(+), 5 deletions(-)
diff --git a/litellm/responses/utils.py b/litellm/responses/utils.py
index 42bd99ccae2..d3418da585d 100644
--- a/litellm/responses/utils.py
+++ b/litellm/responses/utils.py
@@ -285,10 +285,11 @@ class ResponsesAPIRequestUtils:
)
# decode previous_response_id if it's a litellm encoded id
- if "previous_response_id" in non_default_params:
+ previous_response_id: Final = non_default_params.get("previous_response_id")
+ if isinstance(previous_response_id, str):
decoded_previous_response_id: Final = (
ResponsesAPIRequestUtils.decode_previous_response_id_to_original_previous_response_id(
- cast(str, non_default_params["previous_response_id"])
+ previous_response_id
)
)
non_default_params["previous_response_id"] = decoded_previous_response_id
@@ -296,9 +297,8 @@ class ResponsesAPIRequestUtils:
if "metadata" in non_default_params:
from litellm.utils import add_openai_metadata
- converted_metadata: Final = add_openai_metadata(
- cast(Mapping[str, object] | None, non_default_params["metadata"])
- )
+ raw_metadata: Final = non_default_params["metadata"]
+ converted_metadata: Final = add_openai_metadata(raw_metadata if _is_object_dict(raw_metadata) else None)
if converted_metadata is not None:
non_default_params["metadata"] = converted_metadata
else:
From 0a4719697e3e2d8181cf383ba5180fa25cd6c001 Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Sat, 12 Sep 2026 18:28:43 +0000
Subject: [PATCH 006/278] fix(ui): list every provider in the cache leakage
by-model table
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
.../_components/CacheLeakageCard.test.tsx | 6 ++--
.../_components/costOptimizationUtils.test.ts | 36 ++++++-------------
.../_components/costOptimizationUtils.ts | 3 --
3 files changed, 14 insertions(+), 31 deletions(-)
diff --git a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CacheLeakageCard.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CacheLeakageCard.test.tsx
index f320d8e0f97..126011723dd 100644
--- a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CacheLeakageCard.test.tsx
+++ b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/CacheLeakageCard.test.tsx
@@ -122,11 +122,11 @@ describe("CacheLeakageCard", () => {
expect(firstDataRow()).toHaveTextContent("alpha");
});
- it("switches to the model view and lists only Anthropic models", () => {
+ it("switches to the model view and lists models from every provider", () => {
renderWith([
dayWithModels("2026-07-12", {
"claude-sonnet-5": { prompt_tokens: 5000, cache_read_input_tokens: 0 },
- "gpt-4o": { prompt_tokens: 8000, cache_read_input_tokens: 0 },
+ "vertex_ai/gemini-2.5-pro": { prompt_tokens: 8000, cache_read_input_tokens: 2000 },
}),
]);
@@ -134,7 +134,7 @@ describe("CacheLeakageCard", () => {
expect(screen.getByText("Cache leakage by model")).toBeInTheDocument();
expect(screen.getByText("claude-sonnet-5")).toBeInTheDocument();
- expect(screen.queryByText("gpt-4o")).not.toBeInTheDocument();
+ expect(screen.getByText("vertex_ai/gemini-2.5-pro")).toBeInTheDocument();
});
it("shows an empty state when no key used tokens in the range", () => {
diff --git a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/costOptimizationUtils.test.ts b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/costOptimizationUtils.test.ts
index 5d2c48e6440..9c3915c812f 100644
--- a/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/costOptimizationUtils.test.ts
+++ b/ui/litellm-dashboard/src/app/(dashboard)/cost-optimization/_components/costOptimizationUtils.test.ts
@@ -10,7 +10,6 @@ import {
classificationRatePer1kTurns,
computeCacheLeakage,
formatRangeLabel,
- isAnthropicModel,
localIsoDay,
savingsSeriesOf,
toCumulative,
@@ -209,20 +208,21 @@ describe("computeCacheLeakage", () => {
});
describe("computeCacheLeakage by model", () => {
- it("aggregates only Anthropic models and ignores other providers", () => {
+ it("lists every provider's models, not only Anthropic", () => {
const models: RecordInput: ${model.input_cost}/1M tokens
} + {showOutput &&Output: ${model.output_cost}/1M tokens
} + {hasPerSecond &&Output: {formatPerSecondCost(perSecond)}
} + {(model.output_cost_per_second_tiers ?? []).map(({ resolution, cost }) => ( ++ Output ({resolution}): {formatPerSecondCost(cost)} +
+ ))} +
-
- {showTooltip && (
-
- );
-};
-
/**
* A dropdown menu for multiple documentation links.
* Linear-style: Single "Docs" button that expands to show multiple relevant links.
From 726c2bb6df672e07709023f9a3f7350730d71c0f Mon Sep 17 00:00:00 2001
From: mateo
-
- )}
- {content}
- {learnMoreHref && (
-
- {learnMoreText}
-
-
- )}
-
-
- {/* Version Bar */}
-
-
-
- {/* Parameters */}
-
-
-
- {showVersionHistory && (
-
- Version:
-
-
-
-
-
-
-
-
- {versions.map((v) => (
-
- )}
-
-
- ))}
-
-
- {v.id}
-
- {v.changes}
-
-
- {v.author}
- {v.date}
-
-
-
-
- {/* Custom Code Override */}
- Parameters
-Configure {guardrailName} behavior
- -
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
- {useCustomCode && (
-
-
-
-
-
-
- - Replace the built-in guardrail with custom evaluation code -
-