feat - add openai web search

litellm/litellm_core_utils/llm_response_utils/convert_dict_to_response.py

@@ -494,6 +494,7 @@ def convert_to_model_response_object(  # noqa: PLR0915
                         provider_specific_fields=provider_specific_fields,
                         reasoning_content=reasoning_content,
                         thinking_blocks=thinking_blocks,
+                        annotations=choice["message"].get("annotations", None),
                     )
                     finish_reason = choice.get("finish_reason", None)
                 if finish_reason is None:

litellm/types/llms/openai.py

@@ -382,6 +382,28 @@ class ChatCompletionThinkingBlock(TypedDict, total=False):
     cache_control: Optional[Union[dict, ChatCompletionCachedContent]]
 
 
+class ChatCompletionAnnotationURLCitation(TypedDict, total=False):
+    end_index: int
+    """The index of the last character of the URL citation in the message."""
+
+    start_index: int
+    """The index of the first character of the URL citation in the message."""
+
+    title: str
+    """The title of the web resource."""
+
+    url: str
+    """The URL of the web resource."""
+
+
+class ChatCompletionAnnotation(TypedDict, total=False):
+    type: Literal["url_citation"]
+    """The type of the URL citation. Always `url_citation`."""
+
+    url_citation: ChatCompletionAnnotationURLCitation
+    """A URL citation when using web search."""
+
+
 class OpenAIChatCompletionTextObject(TypedDict):
     type: Literal["text"]
     text: str

litellm/types/utils.py

@@ -7,6 +7,7 @@ from typing import Any, Dict, List, Literal, Optional, Tuple, Union
 from aiohttp import FormData
 from openai._models import BaseModel as OpenAIObject
 from openai.types.audio.transcription_create_params import FileTypes  # type: ignore
+from openai.types.chat.chat_completion import ChatCompletion
 from openai.types.completion_usage import (
     CompletionTokensDetails,
     CompletionUsage,
@@ -27,6 +28,7 @@ from ..litellm_core_utils.core_helpers import map_finish_reason
 from .guardrails import GuardrailEventHooks
 from .llms.openai import (
     Batch,
+    ChatCompletionAnnotation,
     ChatCompletionThinkingBlock,
     ChatCompletionToolCallChunk,
     ChatCompletionUsageBlock,
@@ -527,6 +529,7 @@ class Message(OpenAIObject):
     provider_specific_fields: Optional[Dict[str, Any]] = Field(
         default=None, exclude=True
     )
+    annotations: Optional[List[ChatCompletionAnnotation]] = None
 
     def __init__(
         self,
@@ -538,6 +541,7 @@ class Message(OpenAIObject):
         provider_specific_fields: Optional[Dict[str, Any]] = None,
         reasoning_content: Optional[str] = None,
         thinking_blocks: Optional[List[ChatCompletionThinkingBlock]] = None,
+        annotations: Optional[List[ChatCompletionAnnotation]] = None,
         **params,
     ):
         init_values: Dict[str, Any] = {
@@ -566,6 +570,9 @@ class Message(OpenAIObject):
         if thinking_blocks is not None:
             init_values["thinking_blocks"] = thinking_blocks
 
+        if annotations is not None:
+            init_values["annotations"] = annotations
+
         if reasoning_content is not None:
             init_values["reasoning_content"] = reasoning_content