* attempt to implement the passthrough feature
* Formatting and small change
* Fix formatting
* feat: grayswan guardrail overwrite ModelResponse in passthrough mode
* fix missing exception error catching on certain
endpoints
* fix wrong call site
* fix: patch anthropic endpoint internal error on streaming obj
* fix grayswan testcase
* feat: update the violation response to more natural
* Formatting
* move passthrough exception definition to custom_guardrail.
* Enhancement: show whether the blocked at input or output
* update exception name
* fix a typo in testing unit.
---------
Co-authored-by: Xiaohan Fu <xiaohan@grayswan.ai>
The _delete_nested_value_custom function is recursive but has bounded depth
(limited by the number of path segments), preventing infinite recursion.
This is necessary for nested field removal in additional_drop_params.
The nested field removal feature uses a custom implementation with zero
external dependencies, so poetry.lock should match origin/main. The
previous changes were from the intermediate commit that added jsonpath-ng,
which was later removed.
- Change _delete_nested_value_custom parameter type from Dict[str, Any] to Union[Dict[str, Any], List[Any]] to accurately reflect that it handles both dicts and lists
- Add explicit isinstance checks before recursing into list items to ensure type safety
- Prevents potential runtime errors when encountering primitive types in nested structures
Changes litellm_params.additional_drop_params to litellm_params.get("additional_drop_params")
to avoid MyPy error since additional_drop_params is not a defined Pydantic field.
This matches the existing pattern used throughout llm_http_handler.py for accessing
optional fields (e.g., ssl_verify, timeout, etc.).
Add detection for Cerebras's context window exceeded error format:
"Current length is X while limit is Y"
This ensures LiteLLM raises ContextWindowExceededError instead of
generic BadRequestError when Cerebras API calls exceed the model's
context limit, enabling downstream libraries like DSPy to properly
catch and handle these errors for automatic context management.
The 'user' parameter was being ignored when using responses API models
(e.g., model="openai/responses/gpt-4.1") because the model name check
in get_supported_openai_params() didn't account for the "responses/" prefix.
Fix: Normalize the model name by stripping "responses/" prefix before
checking if the model is in the list of supported OpenAI models.
This is a minimal, non-breaking change that:
- Adds 2 lines of code in gpt_transformation.py
- Only affects the parameter support check, not the model variable itself
- Includes unit and integration tests
When OpenAI Responses API returns both text AND tool_calls, the bridge
transformation was emitting is_finished=True after the text message completed,
causing subsequent tool_call chunks to be dropped.
The fix:
- response.output_item.done for messages no longer emits is_finished=True
- Added handler for response.completed to properly signal stream end
* refactor: remove api-key conversion logic for Azure Anthropic
Co-authored-by: Erdem Halil <erdemhalil@users.noreply.github.com>
* fix(passthrough): pass custom_llm_provider to completion_cost for Azure AI Anthropic
The passthrough logging for Anthropic was failing when using Azure AI Anthropic
because the completion_cost function was not receiving the custom_llm_provider
parameter, causing it to fail with "LLM Provider NOT provided" error.
This fix:
- Retrieves custom_llm_provider from logging_obj.model_call_details
- Prepends provider prefix to model name for cost calculation
- Passes both formatted model and custom_llm_provider to completion_cost
- Centralizes provider prefix logic in _create_anthropic_response_logging_payload
This ensures cost calculation works correctly for Azure AI Anthropic requests
with models like azure_ai/claude-sonnet-4-5_gb_20250929.
Co-authored-by: Erdem Halil <erdemhalil@users.noreply.github.com>
* test: add unit tests for Azure AI Anthropic fixes
- Add tests for custom_llm_provider cost calculation in passthrough logging
- Add tests for ProviderConfigManager returning AzureAnthropicMessagesConfig
- Update existing tests to reflect removal of api-key to x-api-key conversion
Co-authored-by: Erdem Halil <erdemhalil@users.noreply.github.com>
---------
Co-authored-by: Erdem Halil <erdemhalil@users.noreply.github.com>
When using litellm with a recent pydantic, the following deprecation warnings are shown
```
litellm/types/llms/anthropic.py:531: PydanticDeprecatedSince20: Support for class-based `config` is deprecated, use ConfigDict instead. Deprecated in Pydantic V2.0 to be removed in V3.0. See Pydantic V2 Migration Guide at https://errors.pydantic.dev/2.12/migration/
class AnthropicResponseContentBlockToolUse(BaseModel):
litellm/types/rag.py:181: PydanticDeprecatedSince20: Support for class-based `config` is deprecated, use ConfigDict instead. Deprecated in Pydantic V2.0 to be removed in V3.0. See Pydantic V2 Migration Guide at https://errors.pydantic.dev/2.12/migration/
class RAGIngestRequest(BaseModel):
```
This aligns with multiple existing places within the litellm codebase that uses ConfigDict to configure pydantic model behavior.
* docs: update Getting Started page with accurate endpoints and fix exception handling
- Update endpoints list to include /responses, /audio, /batches
- Change "Consistent output" to be endpoint-agnostic
- Clarify Response Format title as "OpenAI Chat Completions Format"
- Fix exception handling example: use litellm exceptions instead of deprecated openai.error
- Add model prefix (anthropic/) to example
* docs: reorganize sidebar and improve SDK documentation structure
Sidebar changes:
- Reorder: Python SDK first, then AI Gateway (Proxy)
- Rename "LiteLLM - Getting Started" to "Getting Started"
- Restructure SDK section with Core Functions, Configuration subsections
- Move budget_manager to Guides
- Move sdk_custom_pricing and migration to Extras
- Remove duplicate embedding/async_embedding and embedding/moderation
Content changes:
- Add Response Format section to response_api.md
- Add async aembedding() section to supported_embedding.md
* docs: add deprecation notice for OpenAI Assistants API
OpenAI has deprecated the Assistants API, shutting down on August 26, 2026.
Added warning banner directing users to the Responses API.
* docs: expand Core Functions in SDK sidebar
Add more SDK functions to Core Functions category:
- text_completion()
- image_generation()
- transcription()
- speech()
- Link to "All Supported Endpoints" for complete list
* Rename Sidebar Item
* docs: revert Getting Started label to original
* Rename sidebar label from 'LiteLLM - Getting Started' to 'Getting Started'
