Gemini

The Gemini model provides access to Google’s Gemini models.

Parameters

Parameter	Type	Default	Description
id	str	"gemini-3.5-flash"	The id of the Gemini model to use
name	str	"Gemini"	The name of the model
provider	str	"Google"	The provider of the model
function_declarations	Optional[List[Any]]	None	Function declarations to make available to the model
generation_config	Optional[Any]	None	Generation configuration parameters for the model
safety_settings	Optional[List[Any]]	None	Safety settings to filter content
generative_model_kwargs	Optional[Dict[str, Any]]	None	Additional keyword arguments for the generative model
search	bool	False	Add the Google Search tool so the model can search the web
grounding	bool	False	Enable grounding with Google Search retrieval
grounding_dynamic_threshold	Optional[float]	None	Dynamic retrieval threshold for grounding
url_context	bool	False	Add the URL context tool so the model can read URLs from the prompt
vertexai_search	bool	False	Enable retrieval from a Vertex AI Search datastore
vertexai_search_datastore	Optional[str]	None	Resource name of the Vertex AI Search datastore. Required when vertexai_search is True
parallel_search	bool	False	Enable Parallel web search grounding (Vertex AI only)
parallel_api_key	Optional[str]	None	API key for Parallel web search
parallel_config	Optional[Dict[str, Any]]	None	Custom configuration for Parallel search, passed as custom_configs (e.g. domain filtering)
file_search_store_names	Optional[List[str]]	None	Gemini File Search store names to search
file_search_metadata_filter	Optional[str]	None	Metadata filter applied to File Search results
temperature	Optional[float]	None	Controls randomness in the model’s output
top_p	Optional[float]	None	Controls diversity via nucleus sampling
top_k	Optional[int]	None	Controls diversity via top-k sampling
max_output_tokens	Optional[int]	None	Maximum number of tokens to generate
stop_sequences	Optional[List[str]]	None	Sequences where the model stops generating further tokens
logprobs	Optional[bool]	None	Whether to return log probabilities of the output tokens
presence_penalty	Optional[float]	None	Penalizes new tokens based on whether they appear in the text so far
frequency_penalty	Optional[float]	None	Penalizes new tokens based on their frequency in the text so far
seed	Optional[int]	None	Random seed for deterministic sampling
response_modalities	Optional[List[str]]	None	Output modalities to request: "TEXT", "IMAGE", and/or "AUDIO"
speech_config	Optional[Dict[str, Any]]	None	Speech generation configuration for audio output
cached_content	Optional[Any]	None	Cached content identifier for context caching
thinking_budget	Optional[int]	None	Thinking token budget for Gemini 2.5 models
include_thoughts	Optional[bool]	None	Include thought summaries in the response
thinking_level	Optional[str]	None	Thinking level: "low" or "high"
request_params	Optional[Dict[str, Any]]	None	Additional parameters to include in the request
timeout	Optional[float]	None	Request timeout in seconds
collect_metrics_on_completion	bool	True	Collect token metrics only on the final streaming chunk. Gemini reports cumulative counts per chunk
credentials	Optional[Credentials]	None	Google Cloud credentials for Vertex AI
api_key	Optional[str]	None	The API key for Google AI (defaults to GOOGLE_API_KEY env var)
vertexai	bool	False	Use the Vertex AI API. Also enabled when the GOOGLE_GENAI_USE_VERTEXAI env var is "true"
project_id	Optional[str]	None	Google Cloud project ID for Vertex AI. Falls back to the GOOGLE_CLOUD_PROJECT env var
location	Optional[str]	None	Google Cloud location for Vertex AI. Falls back to the GOOGLE_CLOUD_LOCATION env var
client_params	Optional[Dict[str, Any]]	None	Additional parameters for client configuration
client	Optional[GeminiClient]	None	A pre-configured instance of the Gemini client
model_type	ModelType	ModelType.MODEL	Functional role of this model (MODEL, OUTPUT_MODEL, or PARSER_MODEL). Set by the agent during initialization
supports_native_structured_outputs	bool	True	True if the model supports structured outputs natively
supports_json_schema_outputs	bool	False	True if the model requires a JSON schema for structured outputs
system_prompt	Optional[str]	None	System prompt from the model added to the Agent
instructions	Optional[List[str]]	None	Instructions from the model added to the Agent
tool_message_role	str	"tool"	Role used for tool messages
assistant_message_role	str	"assistant"	Role used for assistant messages
cache_response	bool	False	Cache model responses to avoid redundant API calls during development
cache_ttl	Optional[int]	None	Time-to-live for cached model responses, in seconds. If None, cache never expires
cache_dir	Optional[str]	None	Directory for cached model responses. If None, uses the default cache location
retries	int	0	Number of retries to attempt before raising a ModelProviderError
delay_between_retries	int	1	Delay between retries, in seconds
exponential_backoff	bool	False	If True, the delay between retries is doubled each time
retry_with_guidance	bool	True	Retry the model invocation with a guidance message for known errors avoidable with extra instructions
retry_with_guidance_limit	int	1	Number of times to retry the model invocation with guidance