From cbf9ffec3050e1f97b1a1e78c7cccd14fa700527 Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Wed, 20 May 2026 04:20:54 +0530 Subject: [PATCH] Day 0 support : Gemini 3.5 Flash (#28268) * Add day 0 support for gemini 3.5 flash * Fix pricing * Fix greptile review * Fix failing test * Fix tests * Fix: revert tool removing logic * fix greptile and test --------- Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com> (cherry picked from commit 3c3d131f017af22e09a1b32d033238a3933085dc) --- .../prompt_templates/factory.py | 40 +++- .../llms/vertex_ai/gemini/transformation.py | 4 +- .../vertex_and_google_ai_studio_gemini.py | 128 ++++++++++++- ...odel_prices_and_context_window_backup.json | 178 ++++++++++++++++++ litellm/types/llms/vertex_ai.py | 21 ++- model_prices_and_context_window.json | 178 ++++++++++++++++++ ...test_vertex_and_google_ai_studio_gemini.py | 32 ++++ .../llms/vertex_ai/test_vertex.py | 40 ++-- 8 files changed, 576 insertions(+), 45 deletions(-) diff --git a/litellm/litellm_core_utils/prompt_templates/factory.py b/litellm/litellm_core_utils/prompt_templates/factory.py index d40ca4e359..d39b7a5056 100644 --- a/litellm/litellm_core_utils/prompt_templates/factory.py +++ b/litellm/litellm_core_utils/prompt_templates/factory.py @@ -1233,6 +1233,7 @@ def infer_protocol_value( def _gemini_tool_call_invoke_helper( function_call_params: ChatCompletionToolCallFunctionChunk, + tool_call_id: Optional[str] = None, ) -> Optional[VertexFunctionCall]: name = function_call_params.get("name", "") or "" arguments = function_call_params.get("arguments", "") @@ -1248,6 +1249,10 @@ def _gemini_tool_call_invoke_helper( name=name, args=arguments_dict, ) + if tool_call_id: + clean_id = tool_call_id.split(THOUGHT_SIGNATURE_SEPARATOR, 1)[0] + if clean_id: + function_call["id"] = clean_id return function_call @@ -1384,12 +1389,23 @@ def convert_to_gemini_tool_call_invoke( tool_calls = message.get("tool_calls", None) function_call = message.get("function_call", None) + from litellm.llms.vertex_ai.gemini.vertex_and_google_ai_studio_gemini import ( + VertexGeminiConfig, + ) + + forward_tool_call_id = bool( + model and VertexGeminiConfig._is_gemini_3_or_newer(model) + ) + if tool_calls is not None: for idx, tool in enumerate(tool_calls): if "function" in tool: gemini_function_call: Optional[VertexFunctionCall] = ( _gemini_tool_call_invoke_helper( - function_call_params=tool["function"] + function_call_params=tool["function"], + tool_call_id=( + tool.get("id") if forward_tool_call_id else None + ), ) ) if gemini_function_call is not None: @@ -1429,10 +1445,6 @@ def convert_to_gemini_tool_call_invoke( thought_signature = provider_fields.get("thought_signature") # If no signature found and model is gemini-3, use dummy signature - from litellm.llms.vertex_ai.gemini.vertex_and_google_ai_studio_gemini import ( - VertexGeminiConfig, - ) - if ( not thought_signature and model @@ -1462,6 +1474,7 @@ def convert_to_gemini_tool_call_invoke( def convert_to_gemini_tool_call_result( # noqa: PLR0915 message: Union[ChatCompletionToolMessage, ChatCompletionFunctionMessage], last_message_with_tool_calls: Optional[dict], + model: Optional[str] = None, ) -> Union[VertexPartType, List[VertexPartType]]: """ OpenAI message with a tool result looks like: @@ -1602,6 +1615,21 @@ def convert_to_gemini_tool_call_result( # noqa: PLR0915 ): name = tool.get("function", {}).get("name", "") + # Echo the OpenAI tool_call_id on functionResponse (strip thought-signature suffix). + # Only Gemini 3+ accepts (and returns) an `id` on function_response parts; + # older Gemini models reject the field with a 400. + from litellm.llms.vertex_ai.gemini.vertex_and_google_ai_studio_gemini import ( + VertexGeminiConfig, + ) + + gemini_call_id: Optional[str] = None + if model and VertexGeminiConfig._is_gemini_3_or_newer(model): + raw_tool_call_id = message.get("tool_call_id") + if raw_tool_call_id and isinstance(raw_tool_call_id, str): + stripped_id = raw_tool_call_id.split(THOUGHT_SIGNATURE_SEPARATOR, 1)[0] + if stripped_id: + gemini_call_id = stripped_id + if not name: raise Exception( "Missing corresponding tool call for tool response message. Received - message={}, last_message_with_tool_calls={}".format( @@ -1632,6 +1660,8 @@ def convert_to_gemini_tool_call_result( # noqa: PLR0915 name=name, response=response_data, # type: ignore ) + if gemini_call_id: + _function_response["id"] = gemini_call_id # Create part with function_response, and optionally inline_data for images (Computer Use) _part: VertexPartType = {"function_response": _function_response} diff --git a/litellm/llms/vertex_ai/gemini/transformation.py b/litellm/llms/vertex_ai/gemini/transformation.py index 9afa5dec46..44d31eb1b8 100644 --- a/litellm/llms/vertex_ai/gemini/transformation.py +++ b/litellm/llms/vertex_ai/gemini/transformation.py @@ -607,7 +607,9 @@ def _gemini_convert_messages_with_history( # noqa: PLR0915 and messages[msg_i]["role"] in tool_call_message_roles ): _part = convert_to_gemini_tool_call_result( - messages[msg_i], last_message_with_tool_calls # type: ignore + messages[msg_i], # type: ignore + last_message_with_tool_calls, # type: ignore + model=model, ) msg_i += 1 # Handle both single part and list of parts (for Computer Use with images) diff --git a/litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py b/litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py index 6278de662f..101828cbb5 100644 --- a/litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py +++ b/litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py @@ -280,6 +280,8 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig): - gemini-3-pro-preview - gemini-3-flash - gemini-3-flash-preview (Gemini 3 Flash) + - gemini-3.1-pro-preview, gemini-3.1-flash, gemini-3.1-flash-lite-preview + - gemini-3.5-flash - Any future Gemini 3.x models """ # Check for Gemini 3 models @@ -300,6 +302,7 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig): supported_params = [ "temperature", "top_p", + "top_k", "max_tokens", "max_completion_tokens", "stream", @@ -363,6 +366,66 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig): """ return Tools(googleSearch={}) + @staticmethod + def _search_tool_keys() -> set: + return { + VertexToolName.GOOGLE_SEARCH.value, + VertexToolName.GOOGLE_SEARCH_RETRIEVAL.value, + VertexToolName.ENTERPRISE_WEB_SEARCH.value, + VertexToolName.URL_CONTEXT.value, + "google_search", + "google_search_retrieval", + "enterprise_web_search", + "urlContext", + } + + @classmethod + def _drop_search_tools_mixed_with_functions(cls, optional_params: dict) -> None: + """ + Drop search tools from optional_params when mixed with function declarations + and include_server_side_tool_invocations is not enabled. + + Runs after map_openai_params merges tools and web_search_options so both + code paths (single _map_function call vs split tools + web_search_options) + get the same conflict resolution. + """ + if optional_params.get("include_server_side_tool_invocations"): + return + + tools = optional_params.get("tools") + if not isinstance(tools, list) or not tools: + return + + search_tool_keys = cls._search_tool_keys() + has_function_declarations = any( + isinstance(tool, dict) and tool.get("function_declarations") + for tool in tools + ) + if not has_function_declarations: + return + + has_search_tools = any( + isinstance(tool, dict) and any(key in tool for key in search_tool_keys) + for tool in tools + ) + if not has_search_tools: + return + + verbose_logger.warning( + "Vertex AI does not support mixing function declarations with " + "search tools (googleSearch, enterpriseWebSearch, urlContext, " + "googleSearchRetrieval) in the same request. Dropping search " + "tools and keeping function declarations. To use search tools, " + "send a request without function calling tools." + ) + optional_params["tools"] = [ + tool + for tool in tools + if not ( + isinstance(tool, dict) and any(key in tool for key in search_tool_keys) + ) + ] + def _map_service_tier_param(self, value: str, optional_params: dict) -> None: """ Map OpenAI service_tier (string) to Gemini serviceTier. @@ -884,9 +947,10 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig): GeminiThinkingConfig with thinkingLevel and includeThoughts """ # Check if this is gemini-3-flash which supports MINIMAL thinking level - # Covers gemini-3-flash, gemini-3-flash-preview, gemini-3.1-flash, gemini-3.1-flash-lite-preview, etc. + # Covers gemini-3-flash, gemini-3-flash-preview, gemini-3.1-flash, gemini-3.1-flash-lite-preview, + # gemini-3.5-flash, and any future 3.x-flash variants. is_gemini3flash = model and ( - "gemini-3-flash" in model.lower() or "gemini-3.1-flash" in model.lower() + "flash" in model.lower() and "gemini-3" in model.lower() ) is_gemini31pro = model and ("gemini-3.1-pro-preview" in model.lower()) if reasoning_effort == "minimal": @@ -982,8 +1046,7 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig): # Follow provider defaults unless explicitly opted into legacy behavior. if litellm.enable_gemini_default_thinking_level_low is True: is_gemini3flash = ( - "gemini-3-flash-preview" in model.lower() - or "gemini-3-flash" in model.lower() + "gemini-3" in model.lower() and "flash" in model.lower() ) params["thinkingLevel"] = ( "minimal" if is_gemini3flash else "low" @@ -1077,6 +1140,7 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig): model: str, drop_params: bool, ) -> Dict: + gemini_sampling_params_warned: bool = False for param, value in non_default_params.items(): if param == "temperature": if VertexGeminiConfig._is_gemini_3_or_newer(model): @@ -1086,9 +1150,41 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig): "can cause infinite loops, degraded reasoning performance, and failure on complex tasks. " "Strongly recommended to use temperature = 1.0 (default)." ) + if not gemini_sampling_params_warned: + verbose_logger.warning( + "DeprecationWarning: `temperature`, `top_p`, and `top_k` continue to " + f"function for Gemini 3+ ({model}) but are planned for removal in a " + "future release. Move sampling guidance into the `system` " + "instructions instead." + ) + gemini_sampling_params_warned = True optional_params["temperature"] = value elif param == "top_p": + if ( + VertexGeminiConfig._is_gemini_3_or_newer(model) + and not gemini_sampling_params_warned + ): + verbose_logger.warning( + "DeprecationWarning: `temperature`, `top_p`, and `top_k` continue to " + f"function for Gemini 3+ ({model}) but are planned for removal in a " + "future release. Move sampling guidance into the `system` " + "instructions instead." + ) + gemini_sampling_params_warned = True optional_params["top_p"] = value + elif param == "top_k": + if ( + VertexGeminiConfig._is_gemini_3_or_newer(model) + and not gemini_sampling_params_warned + ): + verbose_logger.warning( + "DeprecationWarning: `temperature`, `top_p`, and `top_k` continue to " + f"function for Gemini 3+ ({model}) but are planned for removal in a " + "future release. Move sampling guidance into the `system` " + "instructions instead." + ) + gemini_sampling_params_warned = True + optional_params["top_k"] = value elif ( param == "stream" and value is True ): # sending stream = False, can cause it to get passed unchecked and raise issues @@ -1139,11 +1235,14 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig): if _tool_choice_value is not None: optional_params["tool_choice"] = _tool_choice_value elif param == "parallel_tool_calls": - if value is False and not ( - drop_params or litellm.drop_params - ): # if drop params is True, then we should just ignore this - self.validate_parallel_tool_calls(value, non_default_params) - else: + tools_list = non_default_params.get( + "tools", non_default_params.get("functions") + ) + num_tools = len(tools_list) if isinstance(tools_list, list) else 0 + # Gemini does not support parallel_tool_calls=False with multiple + # tools. Drop the param instead of failing — Responses API clients + # often send parallel_tool_calls=false by default. + if not (value is False and num_tools > 1): optional_params["parallel_tool_calls"] = value elif param == "seed": optional_params["seed"] = value @@ -1216,6 +1315,8 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig): if "temperature" not in optional_params: optional_params["temperature"] = 1.0 + self._drop_search_tools_mixed_with_functions(optional_params) + return optional_params def get_mapped_special_auth_params(self) -> dict: @@ -1588,6 +1689,10 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig): } # Extract thought signature if present thought_signature = part.get("thoughtSignature") + # Gemini 3.5+ returns a stable `id` per function call to enable + # strict response matching. Preserve it as the OpenAI + # tool_call_id so it can be echoed back unchanged. + gemini_call_id = part["functionCall"].get("id") if is_function_call is True: function_dict: Dict[str, Any] = dict(_function_chunk) @@ -1605,6 +1710,11 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig): "function": _function_chunk, "index": cumulative_tool_call_idx, } + # Gemini 3.5+ returns a stable native `id`; prefer it over + # the synthetic call_ so the same value can be echoed + # back on the matching `functionResponse`. + if gemini_call_id: + _tool_response_chunk["id"] = gemini_call_id # Embed thought signature in ID for OpenAI client compatibility if thought_signature: _tool_response_chunk["provider_specific_fields"] = { # type: ignore diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 543172b838..d5c11822b4 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -15237,6 +15237,64 @@ }, "web_search_billing_unit": "per_query" }, + "vertex_ai/gemini-3.5-flash": { + "cache_read_input_token_cost": 1.5e-07, + "input_cost_per_token": 1.5e-06, + "input_cost_per_audio_token": 1e-06, + "litellm_provider": "vertex_ai", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, + "max_input_tokens": 1048576, + "max_output_tokens": 65535, + "max_pdf_size_mb": 30, + "max_tokens": 65535, + "max_video_length": 1, + "max_videos_per_prompt": 10, + "mode": "chat", + "output_cost_per_reasoning_token": 9e-06, + "output_cost_per_token": 9e-06, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/completions", + "/v1/batch" + ], + "supported_modalities": [ + "text", + "image", + "audio", + "video" + ], + "supported_output_modalities": [ + "text" + ], + "supports_audio_input": true, + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_url_context": true, + "supports_video_input": true, + "supports_vision": true, + "supports_web_search": true, + "supports_native_streaming": true, + "input_cost_per_token_priority": 2.7e-06, + "input_cost_per_audio_token_priority": 1.8e-06, + "output_cost_per_token_priority": 1.62e-05, + "cache_read_input_token_cost_priority": 2.7e-07, + "supports_service_tier": true, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, + "web_search_billing_unit": "per_query" + }, "vertex_ai/gemini-3.1-pro-preview": { "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, @@ -16614,6 +16672,67 @@ }, "web_search_billing_unit": "per_query" }, + "gemini/gemini-3.5-flash": { + "cache_read_input_token_cost": 1.5e-07, + "input_cost_per_audio_token": 1e-06, + "input_cost_per_token": 1.5e-06, + "litellm_provider": "gemini", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, + "max_input_tokens": 1048576, + "max_output_tokens": 65535, + "max_pdf_size_mb": 30, + "max_tokens": 65535, + "max_video_length": 1, + "max_videos_per_prompt": 10, + "mode": "chat", + "output_cost_per_reasoning_token": 9e-06, + "output_cost_per_token": 9e-06, + "rpm": 2000, + "source": "https://ai.google.dev/pricing/gemini-3", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/completions", + "/v1/batch" + ], + "supported_modalities": [ + "text", + "image", + "audio", + "video" + ], + "supported_output_modalities": [ + "text" + ], + "supports_audio_output": false, + "supports_audio_input": true, + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_url_context": true, + "supports_video_input": true, + "supports_vision": true, + "supports_web_search": true, + "supports_native_streaming": true, + "tpm": 800000, + "input_cost_per_token_priority": 2.7e-06, + "input_cost_per_audio_token_priority": 1.8e-06, + "output_cost_per_token_priority": 1.62e-05, + "cache_read_input_token_cost_priority": 2.7e-07, + "supports_service_tier": true, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, + "web_search_billing_unit": "per_query" + }, "gemini/gemini-3.1-pro-preview": { "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, @@ -16799,6 +16918,65 @@ }, "web_search_billing_unit": "per_query" }, + "gemini-3.5-flash": { + "cache_read_input_token_cost": 1.5e-07, + "input_cost_per_audio_token": 1e-06, + "input_cost_per_token": 1.5e-06, + "litellm_provider": "vertex_ai-language-models", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, + "max_input_tokens": 1048576, + "max_output_tokens": 65535, + "max_pdf_size_mb": 30, + "max_tokens": 65535, + "max_video_length": 1, + "max_videos_per_prompt": 10, + "mode": "chat", + "output_cost_per_reasoning_token": 9e-06, + "output_cost_per_token": 9e-06, + "source": "https://ai.google.dev/pricing/gemini-3", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/completions", + "/v1/batch" + ], + "supported_modalities": [ + "text", + "image", + "audio", + "video" + ], + "supported_output_modalities": [ + "text" + ], + "supports_audio_output": false, + "supports_audio_input": true, + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_url_context": true, + "supports_video_input": true, + "supports_vision": true, + "supports_web_search": true, + "supports_native_streaming": true, + "input_cost_per_token_priority": 2.7e-06, + "input_cost_per_audio_token_priority": 1.8e-06, + "output_cost_per_token_priority": 1.62e-05, + "cache_read_input_token_cost_priority": 2.7e-07, + "supports_service_tier": true, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, + "web_search_billing_unit": "per_query" + }, "gemini/gemini-2.5-pro-preview-tts": { "cache_read_input_token_cost": 1.25e-07, "cache_read_input_token_cost_above_200k_tokens": 2.5e-07, diff --git a/litellm/types/llms/vertex_ai.py b/litellm/types/llms/vertex_ai.py index 87bf11a902..b357b64156 100644 --- a/litellm/types/llms/vertex_ai.py +++ b/litellm/types/llms/vertex_ai.py @@ -14,13 +14,19 @@ from litellm.types.llms.openai import EmbeddingInput GeminiEmbeddingInput = Union[EmbeddingInput, List[List[str]]] -class FunctionResponse(TypedDict): - name: str +class FunctionResponse(TypedDict, total=False): + # `id` correlates this response with the originating `functionCall` part. + # Required by Gemini 3.5+ for strict function-calling response matching. + id: str + name: Required[str] response: Optional[dict] -class FunctionCall(TypedDict): - name: str +class FunctionCall(TypedDict, total=False): + # `id` is returned by Gemini 3.5+ to correlate the corresponding + # `functionResponse`. Older Gemini models omit this field. + id: str + name: Required[str] args: Optional[dict] @@ -45,8 +51,11 @@ class PartType(TypedDict, total=False): media_resolution: Literal["low", "medium", "high"] -class HttpxFunctionCall(TypedDict): - name: str +class HttpxFunctionCall(TypedDict, total=False): + # `id` is returned by Gemini 3.5+ to correlate the corresponding + # `functionResponse`. Older Gemini models omit this field. + id: str + name: Required[str] args: dict diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 269e7daecc..a7b83316ba 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -15242,6 +15242,64 @@ }, "web_search_billing_unit": "per_query" }, + "vertex_ai/gemini-3.5-flash": { + "cache_read_input_token_cost": 1.5e-07, + "input_cost_per_token": 1.5e-06, + "input_cost_per_audio_token": 1e-06, + "litellm_provider": "vertex_ai", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, + "max_input_tokens": 1048576, + "max_output_tokens": 65535, + "max_pdf_size_mb": 30, + "max_tokens": 65535, + "max_video_length": 1, + "max_videos_per_prompt": 10, + "mode": "chat", + "output_cost_per_reasoning_token": 9e-06, + "output_cost_per_token": 9e-06, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/completions", + "/v1/batch" + ], + "supported_modalities": [ + "text", + "image", + "audio", + "video" + ], + "supported_output_modalities": [ + "text" + ], + "supports_audio_input": true, + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_url_context": true, + "supports_video_input": true, + "supports_vision": true, + "supports_web_search": true, + "supports_native_streaming": true, + "input_cost_per_token_priority": 2.7e-06, + "input_cost_per_audio_token_priority": 1.8e-06, + "output_cost_per_token_priority": 1.62e-05, + "cache_read_input_token_cost_priority": 2.7e-07, + "supports_service_tier": true, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, + "web_search_billing_unit": "per_query" + }, "vertex_ai/gemini-3.1-pro-preview": { "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, @@ -16619,6 +16677,67 @@ }, "web_search_billing_unit": "per_query" }, + "gemini/gemini-3.5-flash": { + "cache_read_input_token_cost": 1.5e-07, + "input_cost_per_audio_token": 1e-06, + "input_cost_per_token": 1.5e-06, + "litellm_provider": "gemini", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, + "max_input_tokens": 1048576, + "max_output_tokens": 65535, + "max_pdf_size_mb": 30, + "max_tokens": 65535, + "max_video_length": 1, + "max_videos_per_prompt": 10, + "mode": "chat", + "output_cost_per_reasoning_token": 9e-06, + "output_cost_per_token": 9e-06, + "rpm": 2000, + "source": "https://ai.google.dev/pricing/gemini-3", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/completions", + "/v1/batch" + ], + "supported_modalities": [ + "text", + "image", + "audio", + "video" + ], + "supported_output_modalities": [ + "text" + ], + "supports_audio_output": false, + "supports_audio_input": true, + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_url_context": true, + "supports_video_input": true, + "supports_vision": true, + "supports_web_search": true, + "supports_native_streaming": true, + "tpm": 800000, + "input_cost_per_token_priority": 2.7e-06, + "input_cost_per_audio_token_priority": 1.8e-06, + "output_cost_per_token_priority": 1.62e-05, + "cache_read_input_token_cost_priority": 2.7e-07, + "supports_service_tier": true, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, + "web_search_billing_unit": "per_query" + }, "gemini/gemini-3.1-pro-preview": { "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, @@ -16804,6 +16923,65 @@ }, "web_search_billing_unit": "per_query" }, + "gemini-3.5-flash": { + "cache_read_input_token_cost": 1.5e-07, + "input_cost_per_audio_token": 1e-06, + "input_cost_per_token": 1.5e-06, + "litellm_provider": "vertex_ai-language-models", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, + "max_input_tokens": 1048576, + "max_output_tokens": 65535, + "max_pdf_size_mb": 30, + "max_tokens": 65535, + "max_video_length": 1, + "max_videos_per_prompt": 10, + "mode": "chat", + "output_cost_per_reasoning_token": 9e-06, + "output_cost_per_token": 9e-06, + "source": "https://ai.google.dev/pricing/gemini-3", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/completions", + "/v1/batch" + ], + "supported_modalities": [ + "text", + "image", + "audio", + "video" + ], + "supported_output_modalities": [ + "text" + ], + "supports_audio_output": false, + "supports_audio_input": true, + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_url_context": true, + "supports_video_input": true, + "supports_vision": true, + "supports_web_search": true, + "supports_native_streaming": true, + "input_cost_per_token_priority": 2.7e-06, + "input_cost_per_audio_token_priority": 1.8e-06, + "output_cost_per_token_priority": 1.62e-05, + "cache_read_input_token_cost_priority": 2.7e-07, + "supports_service_tier": true, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, + "web_search_billing_unit": "per_query" + }, "gemini/gemini-2.5-pro-preview-tts": { "cache_read_input_token_cost": 1.25e-07, "cache_read_input_token_cost_above_200k_tokens": 2.5e-07, diff --git a/tests/test_litellm/llms/vertex_ai/gemini/test_vertex_and_google_ai_studio_gemini.py b/tests/test_litellm/llms/vertex_ai/gemini/test_vertex_and_google_ai_studio_gemini.py index 353d19b019..59c27f56bb 100644 --- a/tests/test_litellm/llms/vertex_ai/gemini/test_vertex_and_google_ai_studio_gemini.py +++ b/tests/test_litellm/llms/vertex_ai/gemini/test_vertex_and_google_ai_studio_gemini.py @@ -2959,6 +2959,38 @@ def test_vertex_ai_gemini3_tool_combination_no_drop(): assert len(tools) == 3 +def test_vertex_ai_mixed_tools_and_web_search_options_drops_search(): + """ + When function tools and web_search_options are sent separately (Codex-style), + search tools are dropped unless include_server_side_tool_invocations is set. + """ + v = VertexGeminiConfig() + optional_params: dict = {} + non_default_params = { + "tools": [ + { + "type": "function", + "function": {"name": "exec_command", "description": "Run a command"}, + } + ], + "web_search_options": {}, + } + + result = v.map_openai_params( + non_default_params=non_default_params, + optional_params=optional_params, + model="gemini-3.5-flash", + drop_params=True, + ) + + assert not result.get("include_server_side_tool_invocations") + tool_keys = set() + for tool in result.get("tools", []): + tool_keys.update(tool.keys()) + assert "function_declarations" in tool_keys + assert "googleSearch" not in tool_keys + + def test_vertex_ai_openai_web_search_tool_transformation(): """ Test that OpenAI-style web_search and web_search_preview tools are transformed to googleSearch. diff --git a/tests/test_litellm/llms/vertex_ai/test_vertex.py b/tests/test_litellm/llms/vertex_ai/test_vertex.py index 2e9629f95d..5977fab68e 100644 --- a/tests/test_litellm/llms/vertex_ai/test_vertex.py +++ b/tests/test_litellm/llms/vertex_ai/test_vertex.py @@ -1490,39 +1490,31 @@ def test_vertex_parallel_tool_calls_true(): assert "tools" in optional_params -def test_vertex_parallel_tool_calls_false_multiple_tools_error(): +def test_vertex_parallel_tool_calls_false_multiple_tools_dropped(): """ - Test that parallel_tool_calls = False with multiple tools raises UnsupportedParamsError - when drop_params is False. + parallel_tool_calls=False with multiple tools is dropped for Gemini + (unsupported upstream). Request should succeed without the param. """ tools = [ {"type": "function", "function": {"name": "get_weather"}}, {"type": "function", "function": {"name": "get_time"}}, ] - with pytest.raises(litellm.utils.UnsupportedParamsError) as excinfo: - get_optional_params( - model="gemini-1.5-pro", - custom_llm_provider="vertex_ai", - tools=tools, - parallel_tool_calls=False, - ) - assert ( - "`parallel_tool_calls=False` is not supported by Gemini when multiple tools are" - in str(excinfo.value) + optional_params = get_optional_params( + model="gemini-1.5-pro", + custom_llm_provider="vertex_ai", + tools=tools, + parallel_tool_calls=False, ) + assert "parallel_tool_calls" not in optional_params + assert "tools" in optional_params - # works when specified as "functions" - with pytest.raises(litellm.utils.UnsupportedParamsError) as excinfo: - get_optional_params( - model="gemini-1.5-pro", - custom_llm_provider="vertex_ai", - functions=tools, - parallel_tool_calls=False, - ) - assert ( - "`parallel_tool_calls=False` is not supported by Gemini when multiple tools are" - in str(excinfo.value) + optional_params = get_optional_params( + model="gemini-1.5-pro", + custom_llm_provider="vertex_ai", + functions=tools, + parallel_tool_calls=False, ) + assert "parallel_tool_calls" not in optional_params def test_vertex_parallel_tool_calls_false_single_tool():