google-genai 1.14.0__py3-none-any.whl → 1.16.0__py3-none-any.whl

google/genai/_transformers.py

@@ -27,6 +27,7 @@ import time
 import types as builtin_types
 import typing
 from typing import Any, GenericAlias, Optional, Sequence, Union  # type: ignore[attr-defined]
+from ._mcp_utils import mcp_to_gemini_tool
 
 if typing.TYPE_CHECKING:
   import PIL.Image
@@ -47,6 +48,20 @@ else:
   _UNION_TYPES = (typing.Union,)
   from typing_extensions import TypeGuard
 
+if typing.TYPE_CHECKING:
+  from mcp import ClientSession as McpClientSession
+  from mcp.types import Tool as McpTool
+else:
+  McpClientSession: typing.Type = Any
+  McpTool: typing.Type = Any
+  try:
+    from mcp import ClientSession as McpClientSession
+    from mcp.types import Tool as McpTool
+  except ImportError:
+    McpClientSession = None
+    McpTool = None
+
+
 def _resource_name(
     client: _api_client.BaseApiClient,
     resource_name: str,
@@ -211,7 +226,9 @@ def t_extract_models(
     return []
 
 
-def t_caches_model(api_client: _api_client.BaseApiClient, model: str) -> Optional[str]:
+def t_caches_model(
+    api_client: _api_client.BaseApiClient, model: str
+) -> Optional[str]:
   model = t_model(api_client, model)
   if not model:
     return None
@@ -262,10 +279,11 @@ def t_function_response(
     return function_response
   else:
     raise TypeError(
-        f'Could not parse input as FunctionResponse. Unsupported'
+        'Could not parse input as FunctionResponse. Unsupported'
         f' function_response type: {type(function_response)}'
     )
 
+
 def t_function_responses(
     function_responses: Union[
         types.FunctionResponseOrDict,
@@ -361,7 +379,9 @@ def t_part(part: Optional[types.PartUnionDict]) -> types.Part:
 
 
 def t_parts(
-    parts: Optional[Union[list[types.PartUnionDict], types.PartUnionDict, list[types.Part]]],
+    parts: Optional[
+        Union[list[types.PartUnionDict], types.PartUnionDict, list[types.Part]]
+    ],
 ) -> list[types.Part]:
   #
   if parts is None or (isinstance(parts, list) and not parts):
@@ -442,9 +462,7 @@ def t_contents_for_embed(
             if part.text:
               text_parts.append(part.text)
             else:
-              logger.warning(
-                  f'Non-text part found, only returning text parts.'
-              )
+              logger.warning(f'Non-text part found, only returning text parts.')
     return text_parts
   else:
     return transformed_contents
@@ -471,7 +489,9 @@ def t_contents(
   result: list[types.Content] = []
   accumulated_parts: list[types.Part] = []
 
-  def _is_part(part: Union[types.PartUnionDict, Any]) -> TypeGuard[types.PartUnionDict]:
+  def _is_part(
+      part: Union[types.PartUnionDict, Any],
+  ) -> TypeGuard[types.PartUnionDict]:
     if (
         isinstance(part, str)
         or isinstance(part, types.File)
@@ -538,7 +558,7 @@ def t_contents(
         result.append(types.UserContent(parts=content))  # type: ignore[arg-type]
       else:
         result.append(content)
-    elif (_is_part(content)):
+    elif _is_part(content):
       _handle_current_part(result, accumulated_parts, content)
     elif isinstance(content, dict):
       # PactDict is already handled in _is_part
@@ -777,7 +797,9 @@ def _process_enum(
   return types.Schema.model_validate(enum_schema)
 
 
-def _is_type_dict_str_any(origin: Union[types.SchemaUnionDict, Any]) -> TypeGuard[dict[str, Any]]:
+def _is_type_dict_str_any(
+    origin: Union[types.SchemaUnionDict, Any],
+) -> TypeGuard[dict[str, Any]]:
   """Verifies the schema is of type dict[str, Any] for mypy type checking."""
   return isinstance(origin, dict) and all(
       isinstance(key, str) for key in origin
@@ -844,27 +866,32 @@ def t_speech_config(
             prebuilt_voice_config=types.PrebuiltVoiceConfig(voice_name=origin)
         )
     )
-  if (
-      isinstance(origin, dict)
-      and 'voice_config' in origin
-      and origin['voice_config'] is not None
-      and 'prebuilt_voice_config' in origin['voice_config']
-      and origin['voice_config']['prebuilt_voice_config'] is not None
-      and 'voice_name' in origin['voice_config']['prebuilt_voice_config']
-  ):
-    return types.SpeechConfig(
-        voice_config=types.VoiceConfig(
-            prebuilt_voice_config=types.PrebuiltVoiceConfig(
-                voice_name=origin['voice_config']['prebuilt_voice_config'].get(
-                    'voice_name'
-                )
-            )
-        )
-    )
+  if isinstance(origin, dict):
+    return types.SpeechConfig.model_validate(origin)
+
   raise ValueError(f'Unsupported speechConfig type: {type(origin)}')
 
 
-def t_tool(client: _api_client.BaseApiClient, origin: Any) -> Optional[Union[types.Tool, Any]]:
+def t_live_speech_config(
+    client: _api_client.BaseApiClient,
+    origin: types.SpeechConfigOrDict,
+) -> Optional[types.SpeechConfig]:
+  if isinstance(origin, types.SpeechConfig):
+    speech_config = origin
+  if isinstance(origin, dict):
+    speech_config = types.SpeechConfig.model_validate(origin)
+
+  if speech_config.multi_speaker_voice_config is not None:
+    raise ValueError(
+        'multi_speaker_voice_config is not supported in the live API.'
+    )
+
+  return speech_config
+
+
+def t_tool(
+    client: _api_client.BaseApiClient, origin: Any
+) -> Optional[Union[types.Tool, Any]]:
   if not origin:
     return None
   if inspect.isfunction(origin) or inspect.ismethod(origin):
@@ -875,13 +902,14 @@ def t_tool(client: _api_client.BaseApiClient, origin: Any) -> Optional[Union[typ
             )
         ]
     )
+  elif McpTool is not None and isinstance(origin, McpTool):
+    return mcp_to_gemini_tool(origin)
   elif isinstance(origin, dict):
     return types.Tool.model_validate(origin)
   else:
     return origin
 
 
-# Only support functions now.
 def t_tools(
     client: _api_client.BaseApiClient, origin: list[Any]
 ) -> list[types.Tool]:
@@ -911,7 +939,9 @@ def t_cached_content_name(client: _api_client.BaseApiClient, name: str) -> str:
   return _resource_name(client, name, collection_identifier='cachedContents')
 
 
-def t_batch_job_source(client: _api_client.BaseApiClient, src: str) -> types.BatchJobSource:
+def t_batch_job_source(
+    client: _api_client.BaseApiClient, src: str
+) -> types.BatchJobSource:
   if src.startswith('gs://'):
     return types.BatchJobSource(
         format='jsonl',
@@ -926,7 +956,9 @@ def t_batch_job_source(client: _api_client.BaseApiClient, src: str) -> types.Bat
     raise ValueError(f'Unsupported source: {src}')
 
 
-def t_batch_job_destination(client: _api_client.BaseApiClient, dest: str) -> types.BatchJobDestination:
+def t_batch_job_destination(
+    client: _api_client.BaseApiClient, dest: str
+) -> types.BatchJobDestination:
   if dest.startswith('gs://'):
     return types.BatchJobDestination(
         format='jsonl',
@@ -960,7 +992,9 @@ LRO_POLLING_TIMEOUT_SECONDS = 900.0
 LRO_POLLING_MULTIPLIER = 1.5
 
 
-def t_resolve_operation(api_client: _api_client.BaseApiClient, struct: dict[str, Any]) -> Any:
+def t_resolve_operation(
+    api_client: _api_client.BaseApiClient, struct: dict[str, Any]
+) -> Any:
   if (name := struct.get('name')) and '/operations/' in name:
     operation: dict[str, Any] = struct
     total_seconds = 0.0
@@ -1058,8 +1092,7 @@ def t_content_strict(content: types.ContentOrDict) -> types.Content:
     return content
   else:
     raise ValueError(
-        f'Could not convert input (type "{type(content)}") to '
-        '`types.Content`'
+        f'Could not convert input (type "{type(content)}") to `types.Content`'
     )
 
 

google/genai/caches.py

@@ -30,6 +30,42 @@ from .pagers import AsyncPager, Pager
 logger = logging.getLogger('google_genai.caches')
 
 
+def _VideoMetadata_to_mldev(
+    api_client: BaseApiClient,
+    from_object: Union[dict[str, Any], object],
+    parent_object: Optional[dict[str, Any]] = None,
+) -> dict[str, Any]:
+  to_object: dict[str, Any] = {}
+  if getv(from_object, ['fps']) is not None:
+    setv(to_object, ['fps'], getv(from_object, ['fps']))
+
+  if getv(from_object, ['end_offset']) is not None:
+    setv(to_object, ['endOffset'], getv(from_object, ['end_offset']))
+
+  if getv(from_object, ['start_offset']) is not None:
+    setv(to_object, ['startOffset'], getv(from_object, ['start_offset']))
+
+  return to_object
+
+
+def _Blob_to_mldev(
+    api_client: BaseApiClient,
+    from_object: Union[dict[str, Any], object],
+    parent_object: Optional[dict[str, Any]] = None,
+) -> dict[str, Any]:
+  to_object: dict[str, Any] = {}
+  if getv(from_object, ['display_name']) is not None:
+    raise ValueError('display_name parameter is not supported in Gemini API.')
+
+  if getv(from_object, ['data']) is not None:
+    setv(to_object, ['data'], getv(from_object, ['data']))
+
+  if getv(from_object, ['mime_type']) is not None:
+    setv(to_object, ['mimeType'], getv(from_object, ['mime_type']))
+
+  return to_object
+
+
 def _Part_to_mldev(
     api_client: BaseApiClient,
     from_object: Union[dict[str, Any], object],
@@ -37,11 +73,26 @@ def _Part_to_mldev(
 ) -> dict[str, Any]:
   to_object: dict[str, Any] = {}
   if getv(from_object, ['video_metadata']) is not None:
-    raise ValueError('video_metadata parameter is not supported in Gemini API.')
+    setv(
+        to_object,
+        ['videoMetadata'],
+        _VideoMetadata_to_mldev(
+            api_client, getv(from_object, ['video_metadata']), to_object
+        ),
+    )
 
   if getv(from_object, ['thought']) is not None:
     setv(to_object, ['thought'], getv(from_object, ['thought']))
 
+  if getv(from_object, ['inline_data']) is not None:
+    setv(
+        to_object,
+        ['inlineData'],
+        _Blob_to_mldev(
+            api_client, getv(from_object, ['inline_data']), to_object
+        ),
+    )
+
   if getv(from_object, ['code_execution_result']) is not None:
     setv(
         to_object,
@@ -65,9 +116,6 @@ def _Part_to_mldev(
         getv(from_object, ['function_response']),
     )
 
-  if getv(from_object, ['inline_data']) is not None:
-    setv(to_object, ['inlineData'], getv(from_object, ['inline_data']))
-
   if getv(from_object, ['text']) is not None:
     setv(to_object, ['text'], getv(from_object, ['text']))
 
@@ -96,12 +144,59 @@ def _Content_to_mldev(
   return to_object
 
 
+def _FunctionDeclaration_to_mldev(
+    api_client: BaseApiClient,
+    from_object: Union[dict[str, Any], object],
+    parent_object: Optional[dict[str, Any]] = None,
+) -> dict[str, Any]:
+  to_object: dict[str, Any] = {}
+  if getv(from_object, ['behavior']) is not None:
+    setv(to_object, ['behavior'], getv(from_object, ['behavior']))
+
+  if getv(from_object, ['description']) is not None:
+    setv(to_object, ['description'], getv(from_object, ['description']))
+
+  if getv(from_object, ['name']) is not None:
+    setv(to_object, ['name'], getv(from_object, ['name']))
+
+  if getv(from_object, ['parameters']) is not None:
+    setv(to_object, ['parameters'], getv(from_object, ['parameters']))
+
+  if getv(from_object, ['response']) is not None:
+    setv(to_object, ['response'], getv(from_object, ['response']))
+
+  return to_object
+
+
+def _Interval_to_mldev(
+    api_client: BaseApiClient,
+    from_object: Union[dict[str, Any], object],
+    parent_object: Optional[dict[str, Any]] = None,
+) -> dict[str, Any]:
+  to_object: dict[str, Any] = {}
+  if getv(from_object, ['start_time']) is not None:
+    setv(to_object, ['startTime'], getv(from_object, ['start_time']))
+
+  if getv(from_object, ['end_time']) is not None:
+    setv(to_object, ['endTime'], getv(from_object, ['end_time']))
+
+  return to_object
+
+
 def _GoogleSearch_to_mldev(
     api_client: BaseApiClient,
     from_object: Union[dict[str, Any], object],
     parent_object: Optional[dict[str, Any]] = None,
 ) -> dict[str, Any]:
   to_object: dict[str, Any] = {}
+  if getv(from_object, ['time_range_filter']) is not None:
+    setv(
+        to_object,
+        ['timeRangeFilter'],
+        _Interval_to_mldev(
+            api_client, getv(from_object, ['time_range_filter']), to_object
+        ),
+    )
 
   return to_object
 
@@ -214,12 +309,32 @@ def _GoogleMaps_to_mldev(
   return to_object
 
 
+def _UrlContext_to_mldev(
+    api_client: BaseApiClient,
+    from_object: Union[dict[str, Any], object],
+    parent_object: Optional[dict[str, Any]] = None,
+) -> dict[str, Any]:
+  to_object: dict[str, Any] = {}
+
+  return to_object
+
+
 def _Tool_to_mldev(
     api_client: BaseApiClient,
     from_object: Union[dict[str, Any], object],
     parent_object: Optional[dict[str, Any]] = None,
 ) -> dict[str, Any]:
   to_object: dict[str, Any] = {}
+  if getv(from_object, ['function_declarations']) is not None:
+    setv(
+        to_object,
+        ['functionDeclarations'],
+        [
+            _FunctionDeclaration_to_mldev(api_client, item, to_object)
+            for item in getv(from_object, ['function_declarations'])
+        ],
+    )
+
   if getv(from_object, ['retrieval']) is not None:
     raise ValueError('retrieval parameter is not supported in Gemini API.')
 
@@ -251,16 +366,18 @@ def _Tool_to_mldev(
   if getv(from_object, ['google_maps']) is not None:
     raise ValueError('google_maps parameter is not supported in Gemini API.')
 
-  if getv(from_object, ['code_execution']) is not None:
-    setv(to_object, ['codeExecution'], getv(from_object, ['code_execution']))
-
-  if getv(from_object, ['function_declarations']) is not None:
+  if getv(from_object, ['url_context']) is not None:
     setv(
         to_object,
-        ['functionDeclarations'],
-        getv(from_object, ['function_declarations']),
+        ['urlContext'],
+        _UrlContext_to_mldev(
+            api_client, getv(from_object, ['url_context']), to_object
+        ),
     )
 
+  if getv(from_object, ['code_execution']) is not None:
+    setv(to_object, ['codeExecution'], getv(from_object, ['code_execution']))
+
   return to_object
 
 
@@ -290,10 +407,10 @@ def _LatLng_to_mldev(
 ) -> dict[str, Any]:
   to_object: dict[str, Any] = {}
   if getv(from_object, ['latitude']) is not None:
-    raise ValueError('latitude parameter is not supported in Gemini API.')
+    setv(to_object, ['latitude'], getv(from_object, ['latitude']))
 
   if getv(from_object, ['longitude']) is not None:
-    raise ValueError('longitude parameter is not supported in Gemini API.')
+    setv(to_object, ['longitude'], getv(from_object, ['longitude']))
 
   return to_object
 
@@ -305,7 +422,11 @@ def _RetrievalConfig_to_mldev(
 ) -> dict[str, Any]:
   to_object: dict[str, Any] = {}
   if getv(from_object, ['lat_lng']) is not None:
-    raise ValueError('lat_lng parameter is not supported in Gemini API.')
+    setv(
+        to_object,
+        ['latLng'],
+        _LatLng_to_mldev(api_client, getv(from_object, ['lat_lng']), to_object),
+    )
 
   return to_object
 
@@ -328,8 +449,12 @@ def _ToolConfig_to_mldev(
     )
 
   if getv(from_object, ['retrieval_config']) is not None:
-    raise ValueError(
-        'retrieval_config parameter is not supported in Gemini API.'
+    setv(
+        to_object,
+        ['retrievalConfig'],
+        _RetrievalConfig_to_mldev(
+            api_client, getv(from_object, ['retrieval_config']), to_object
+        ),
     )
 
   return to_object
@@ -393,6 +518,9 @@ def _CreateCachedContentConfig_to_mldev(
         ),
     )
 
+  if getv(from_object, ['kms_key_name']) is not None:
+    raise ValueError('kms_key_name parameter is not supported in Gemini API.')
+
   return to_object
 
 
@@ -540,6 +668,42 @@ def _ListCachedContentsParameters_to_mldev(
   return to_object
 
 
+def _VideoMetadata_to_vertex(
+    api_client: BaseApiClient,
+    from_object: Union[dict[str, Any], object],
+    parent_object: Optional[dict[str, Any]] = None,
+) -> dict[str, Any]:
+  to_object: dict[str, Any] = {}
+  if getv(from_object, ['fps']) is not None:
+    setv(to_object, ['fps'], getv(from_object, ['fps']))
+
+  if getv(from_object, ['end_offset']) is not None:
+    setv(to_object, ['endOffset'], getv(from_object, ['end_offset']))
+
+  if getv(from_object, ['start_offset']) is not None:
+    setv(to_object, ['startOffset'], getv(from_object, ['start_offset']))
+
+  return to_object
+
+
+def _Blob_to_vertex(
+    api_client: BaseApiClient,
+    from_object: Union[dict[str, Any], object],
+    parent_object: Optional[dict[str, Any]] = None,
+) -> dict[str, Any]:
+  to_object: dict[str, Any] = {}
+  if getv(from_object, ['display_name']) is not None:
+    setv(to_object, ['displayName'], getv(from_object, ['display_name']))
+
+  if getv(from_object, ['data']) is not None:
+    setv(to_object, ['data'], getv(from_object, ['data']))
+
+  if getv(from_object, ['mime_type']) is not None:
+    setv(to_object, ['mimeType'], getv(from_object, ['mime_type']))
+
+  return to_object
+
+
 def _Part_to_vertex(
     api_client: BaseApiClient,
     from_object: Union[dict[str, Any], object],
@@ -547,11 +711,26 @@ def _Part_to_vertex(
 ) -> dict[str, Any]:
   to_object: dict[str, Any] = {}
   if getv(from_object, ['video_metadata']) is not None:
-    setv(to_object, ['videoMetadata'], getv(from_object, ['video_metadata']))
+    setv(
+        to_object,
+        ['videoMetadata'],
+        _VideoMetadata_to_vertex(
+            api_client, getv(from_object, ['video_metadata']), to_object
+        ),
+    )
 
   if getv(from_object, ['thought']) is not None:
     setv(to_object, ['thought'], getv(from_object, ['thought']))
 
+  if getv(from_object, ['inline_data']) is not None:
+    setv(
+        to_object,
+        ['inlineData'],
+        _Blob_to_vertex(
+            api_client, getv(from_object, ['inline_data']), to_object
+        ),
+    )
+
   if getv(from_object, ['code_execution_result']) is not None:
     setv(
         to_object,
@@ -575,9 +754,6 @@ def _Part_to_vertex(
         getv(from_object, ['function_response']),
     )
 
-  if getv(from_object, ['inline_data']) is not None:
-    setv(to_object, ['inlineData'], getv(from_object, ['inline_data']))
-
   if getv(from_object, ['text']) is not None:
     setv(to_object, ['text'], getv(from_object, ['text']))
 
@@ -606,12 +782,59 @@ def _Content_to_vertex(
   return to_object
 
 
+def _FunctionDeclaration_to_vertex(
+    api_client: BaseApiClient,
+    from_object: Union[dict[str, Any], object],
+    parent_object: Optional[dict[str, Any]] = None,
+) -> dict[str, Any]:
+  to_object: dict[str, Any] = {}
+  if getv(from_object, ['behavior']) is not None:
+    raise ValueError('behavior parameter is not supported in Vertex AI.')
+
+  if getv(from_object, ['description']) is not None:
+    setv(to_object, ['description'], getv(from_object, ['description']))
+
+  if getv(from_object, ['name']) is not None:
+    setv(to_object, ['name'], getv(from_object, ['name']))
+
+  if getv(from_object, ['parameters']) is not None:
+    setv(to_object, ['parameters'], getv(from_object, ['parameters']))
+
+  if getv(from_object, ['response']) is not None:
+    setv(to_object, ['response'], getv(from_object, ['response']))
+
+  return to_object
+
+
+def _Interval_to_vertex(
+    api_client: BaseApiClient,
+    from_object: Union[dict[str, Any], object],
+    parent_object: Optional[dict[str, Any]] = None,
+) -> dict[str, Any]:
+  to_object: dict[str, Any] = {}
+  if getv(from_object, ['start_time']) is not None:
+    setv(to_object, ['startTime'], getv(from_object, ['start_time']))
+
+  if getv(from_object, ['end_time']) is not None:
+    setv(to_object, ['endTime'], getv(from_object, ['end_time']))
+
+  return to_object
+
+
 def _GoogleSearch_to_vertex(
     api_client: BaseApiClient,
     from_object: Union[dict[str, Any], object],
     parent_object: Optional[dict[str, Any]] = None,
 ) -> dict[str, Any]:
   to_object: dict[str, Any] = {}
+  if getv(from_object, ['time_range_filter']) is not None:
+    setv(
+        to_object,
+        ['timeRangeFilter'],
+        _Interval_to_vertex(
+            api_client, getv(from_object, ['time_range_filter']), to_object
+        ),
+    )
 
   return to_object
 
@@ -736,12 +959,32 @@ def _GoogleMaps_to_vertex(
   return to_object
 
 
+def _UrlContext_to_vertex(
+    api_client: BaseApiClient,
+    from_object: Union[dict[str, Any], object],
+    parent_object: Optional[dict[str, Any]] = None,
+) -> dict[str, Any]:
+  to_object: dict[str, Any] = {}
+
+  return to_object
+
+
 def _Tool_to_vertex(
     api_client: BaseApiClient,
     from_object: Union[dict[str, Any], object],
     parent_object: Optional[dict[str, Any]] = None,
 ) -> dict[str, Any]:
   to_object: dict[str, Any] = {}
+  if getv(from_object, ['function_declarations']) is not None:
+    setv(
+        to_object,
+        ['functionDeclarations'],
+        [
+            _FunctionDeclaration_to_vertex(api_client, item, to_object)
+            for item in getv(from_object, ['function_declarations'])
+        ],
+    )
+
   if getv(from_object, ['retrieval']) is not None:
     setv(to_object, ['retrieval'], getv(from_object, ['retrieval']))
 
@@ -783,16 +1026,12 @@ def _Tool_to_vertex(
         ),
     )
 
+  if getv(from_object, ['url_context']) is not None:
+    raise ValueError('url_context parameter is not supported in Vertex AI.')
+
   if getv(from_object, ['code_execution']) is not None:
     setv(to_object, ['codeExecution'], getv(from_object, ['code_execution']))
 
-  if getv(from_object, ['function_declarations']) is not None:
-    setv(
-        to_object,
-        ['functionDeclarations'],
-        getv(from_object, ['function_declarations']),
-    )
-
   return to_object
 
 
@@ -935,6 +1174,13 @@ def _CreateCachedContentConfig_to_vertex(
         ),
     )
 
+  if getv(from_object, ['kms_key_name']) is not None:
+    setv(
+        parent_object,
+        ['encryption_spec', 'kmsKeyName'],
+        getv(from_object, ['kms_key_name']),
+    )
+
   return to_object
 
 
@@ -1082,6 +1328,11 @@ def _ListCachedContentsParameters_to_vertex(
   return to_object
 
 
+def _Behavior_to_vertex_enum_validate(enum_value: Any) -> None:
+  if enum_value in set(['UNSPECIFIED', 'BLOCKING', 'NON_BLOCKING']):
+    raise ValueError(f'{enum_value} enum value is not supported in Vertex AI.')
+
+
 def _CachedContent_from_mldev(
     api_client: BaseApiClient,
     from_object: Union[dict[str, Any], object],

google/genai/chats.py

@@ -1,4 +1,4 @@
-# Copyright 2024 Google LLC
+# Copyright 2025 Google LLC
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.