Deep Researcher

Modules concerning the deep researchers used in Gen AI applications.

GLOpenDeepResearcher(profile='GPTR-QUICK', api_key=None, base_url=None, prompt_builder=None, retry_config=None, tools=None, api_config=None, streamable=True)

Bases: BaseDeepResearcher

A deep researcher that interacts with GL Open Deep Research API.

This class extends the BaseDeepResearcher and interacts with GL Open Deep Research API using the gl_odr client library.

Examples:

deep_researcher = GLOpenDeepResearcher(profile="GPTR-QUICK")
output = await deep_researcher.research(query=query, event_emitter=event_emitter)

Attributes:

Name	Type	Description
profile	str	The research profile to use.
client_params	dict[str, Any]	The parameters to be passed to the client.
prompt_builder	PromptBuilder	The prompt builder to be used to format the inputs.
retry_config	RetryConfig	The retry configuration to be used to conduct the deep research.
streamable	bool	Whether the deep research will be streamed if an event emitter is provided.

Initializes a new instance of the GLOpenDeepResearcher class.

Parameters:

Name	Type	Description	Default
profile	str	The research profile to use. Can be a default profile or a custom profile. Available default profiles include: 1. TONGYI-OPENROUTER 2. TONGYI 3. GPTR-QUICK 4. GPTR-DEEP Custom profiles can be created by referring to https://gdplabs.gitbook.io/gl-open-deepresearch/developers-guide/customization/profiles	'GPTR-QUICK'
api_key	str | None	The API key for GL Open Deep Research. Defaults to None (reads from GLODR_API_KEY env var).	None
base_url	str | None	The base URL for the API. Defaults to None (uses DEFAULT_BASE_URL).	None
prompt_builder	PromptBuilder | None	The prompt builder to format inputs. Defaults to None, in which case a default prompt builder will be used.	None
retry_config	RetryConfig | None	The retry configuration. Defaults to a retry configuration with a timeout of DEFAULT_TIMEOUT.	None
tools	list[NativeTool] | None	The MCP server tools to use. Only MCP server native tools are supported. Defaults to None, in which case the profile's default mcp_config will be used.	None
api_config	dict[str, Any] | None	The GL ODR-specific API configuration for external REST endpoints. Defaults to None, in which case the profile's default api_config will be used.	None
streamable	bool	Whether to stream if event emitter is provided. Defaults to True.	True

Raises:

Type	Description
ImportError	If the gl_odr package is not installed.
ValueError	If any tool is not an MCP server native tool.

get_profile(name) async

Fetch the profile tool config format.

Parameters:

Name	Type	Description	Default
name	str	The profile name to fetch.	required

Returns:

Name	Type	Description
Any	ProfileResponse	The profile response containing api_config and mcp_config formats.

research(query, history=None, extra_contents=None, hyperparameters=None, event_emitter=None, **kwargs) async

Conducts deep research on the given query.

Parameters:

Name	Type	Description	Default
query	str	The query to research.	required
history	list[Message] | None	The conversation history. Defaults to None.	None
extra_contents	list[MessageContent] | None	Extra contents to include. Only text contents are supported. Defaults to None.	None
hyperparameters	dict[str, Any] | None	Additional hyperparameters. Defaults to None.	None
event_emitter	EventEmitter | None	The event emitter for streaming. Defaults to None.	None
**kwargs	Any	Additional keyword arguments.	{}

Returns:

Name	Type	Description
LMOutput	LMOutput	The research output.

Raises:

Type	Description
ValueError	If extra_contents contains non-text content.

GoogleDeepResearcher(agent_name='deep-research-pro-preview-12-2025', api_key=None, prompt_builder=None, retry_config=None, data_stores=None, streamable=True, show_progress=True, status_check_interval=DEFAULT_STATUS_CHECK_INTERVAL)

Bases: BaseDeepResearcher

A deep researcher that interacts with Google deep research agents.

This class extends the BaseDeepResearcher and interacts with Google deep research agents.

Examples:

deep_researcher = GoogleDeepResearcher()
output = await deep_researcher.research(query=query, event_emitter=event_emitter)

Notes

As of March 2026, the Google deep research API has an intermittent missing final output issue. This issue only happens when the stream=True parameter is sent to Google's interactions API. As a workaround, the show_progress parameter can be set to False to avoid the issue. This will ensure that the final output exists, with the cost of not being able to stream the progress.

Attributes:

Name	Type	Description
agent_name	str	The name of the Google deep research agent to use.
client_params	dict[str, Any]	The parameters to be passed to the Google client.
prompt_builder	PromptBuilder	The prompt builder to be used to format the inputs.
retry_config	RetryConfig	The retry configuration to be used to conduct the deep research.
streamable	bool	Whether the deep research will be streamed if an event emitter is provided.

Initializes a new instance of the GoogleDeepResearcher class.

Parameters:

Name	Type	Description	Default
agent_name	str	The name of the Google deep research agent to use. Defaults to "deep-research-pro-preview-12-2025".	'deep-research-pro-preview-12-2025'
api_key	str | None	The API key to use for the Google deep research model. Defaults to None.	None
prompt_builder	PromptBuilder | None	The prompt builder to be used to format the inputs. Defaults to None, in which case a default prompt builder will be used.	None
retry_config	RetryConfig | None	The retry configuration to conduct the deep research. Defaults to a retry configuration with a timeout of DEFAULT_TIMEOUT.	None
data_stores	list[AttachmentStore] | None	The data stores to be used as knowledge bases for the Google deep research agent. Defaults to None.	None
streamable	bool	Whether the deep research will be streamed if an event emitter is provided. Defaults to True.	True
show_progress	bool	Whether to show the progress of the deep research. Defaults to True.	True
status_check_interval	float	The interval to check the status of the deep research. Used only when show_progress is False. Defaults to DEFAULT_STATUS_CHECK_INTERVAL.	DEFAULT_STATUS_CHECK_INTERVAL

Raises:

Type	Description
ImportError	If the google.genai package is not installed.
ValueError	If the agent name is not a valid Google deep research agent.

research(query, history=None, extra_contents=None, hyperparameters=None, event_emitter=None, **kwargs) async

Researches the context.

Parameters:

Name	Type	Description	Default
query	str	The query to research.	required
history	list[Message] | None	The conversation history to be considered in the research. Defaults to None.	None
extra_contents	list[MessageContent] | None	A list of extra contents to be included as inputs. Defaults to None.	None
hyperparameters	dict[str, Any] | None	The hyperparameters to be passed to the language model. Defaults to None.	None
event_emitter	EventEmitter | None	The event emitter to be used to stream the research. Defaults to None.	None
**kwargs	Any	Additional keyword arguments to be passed to the research method.	{}

Returns:

Name	Type	Description
LMOutput	LMOutput	The output from the language model.

OpenAIDeepResearcher(model_name='o4-mini-deep-research', api_key=None, prompt_builder=None, retry_config=None, tools=None, streamable=True)

Bases: BaseDeepResearcher

A deep researcher that interacts with OpenAI deep research models.

This class extends the BaseDeepResearcher and interacts with OpenAI deep research models.

Examples:

deep_researcher = OpenAIDeepResearcher()
output = await deep_researcher.research(query=query, event_emitter=event_emitter)

Attributes:

Name	Type	Description
lm_request_processor	LMRequestProcessor	The LM request processor containing the OpenAI deep research model.
streamable	bool	Whether the deep research will be streamed if an event emitter is provided.

Initializes a new instance of the OpenAIDeepResearcher class.

Parameters:

Name	Type	Description	Default
model_name	str	The name of the OpenAI deep research model to use. Defaults to "o4-mini-deep-research".	'o4-mini-deep-research'
api_key	str | None	The API key to use for the OpenAI deep research model. Defaults to None.	None
prompt_builder	PromptBuilder | None	The prompt builder to be used to format the inputs. Defaults to None, in which case a default prompt builder will be used.	None
retry_config	RetryConfig | None	The retry configuration to conduct the deep research. Defaults to a retry configuration with a timeout of DEFAULT_TIMEOUT.	None
tools	list[NativeTool] | None	The tools to be used to conduct the deep research. Defaults to None.	None
streamable	bool	Whether the deep research will be streamed if an event emitter is provided. Defaults to True.	True

Raises:

Type	Description
ImportError	If the openai package is not installed.
ValueError	If the model name is not a valid OpenAI deep research model.

research(query, history=None, extra_contents=None, hyperparameters=None, event_emitter=None, **kwargs) async

Researches the context.

Parameters:

Name	Type	Description	Default
query	str	The query to research.	required
history	list[Message] | None	The conversation history to be considered in the research. Defaults to None.	None
extra_contents	list[MessageContent] | None	A list of extra contents to be included as inputs. Defaults to None.	None
hyperparameters	dict[str, Any] | None	The hyperparameters to be passed to the language model. Defaults to None.	None
event_emitter	EventEmitter | None	The event emitter to be used to stream the research. Defaults to None.	None
**kwargs	Any	Additional keyword arguments to be passed to the research method.	{}

Returns:

Name	Type	Description
LMOutput	LMOutput	The output from the language model.

ParallelDeepResearcher(processor_name='ultra', api_key=None, prompt_builder=None, retry_config=None, tools=None, event_idle_timeout=DEFAULT_EVENT_IDLE_TIMEOUT, streamable=True)

Bases: BaseDeepResearcher

A deep researcher that interacts with Parallel deep research processors.

This class extends the BaseDeepResearcher and interacts with Parallel deep research processors.

Examples:

deep_researcher = ParallelDeepResearcher()
output = await deep_researcher.research(query=query, event_emitter=event_emitter)

Attributes:

Name	Type	Description
processor_name	str	The name of the Parallel deep research processor to use.
client_params	dict[str, Any]	The parameters to be passed to the Parallel client.
prompt_builder	PromptBuilder	The prompt builder to be used to format the inputs.
retry_config	RetryConfig	The retry configuration to be used to conduct the deep research.
event_idle_timeout	float	The maximum seconds to wait for a new event before raising a TimeoutError.
streamable	bool	Whether the deep research will be streamed if an event emitter is provided.

Initializes a new instance of the ParallelDeepResearcher class.

Parameters:

Name	Type	Description	Default
processor_name	str	The name of the Parallel deep research processor to use. Defaults to "ultra".	'ultra'
api_key	str | None	The API key to use for the Parallel deep research processor. Defaults to None.	None
prompt_builder	PromptBuilder | None	The prompt builder to be used to format the inputs. Defaults to None, in which case a default prompt builder will be used.	None
retry_config	RetryConfig | None	The retry configuration to conduct the deep research. Defaults to a retry configuration with a timeout of DEFAULT_TIMEOUT.	None
tools	list[NativeTool] | None	The tools to be used to conduct the deep research. Defaults to None.	None
event_idle_timeout	float	The maximum seconds to wait for a new event from the stream before raising a TimeoutError. Defaults to DEFAULT_EVENT_IDLE_TIMEOUT.	DEFAULT_EVENT_IDLE_TIMEOUT
streamable	bool	Whether the deep research will be streamed if an event emitter is provided. Defaults to True.	True

research(query, history=None, extra_contents=None, hyperparameters=None, event_emitter=None, **kwargs) async

Conducts the deep research operation.

Parameters:

Name	Type	Description	Default
query	str	The query to research.	required
history	list[Message] | None	The conversation history to be considered in the research. Defaults to None.	None
extra_contents	list[MessageContent] | None	A list of extra contents to be included as inputs. Defaults to None.	None
hyperparameters	dict[str, Any] | None	The hyperparameters to be passed to the language model. Defaults to None.	None
event_emitter	EventEmitter | None	The event emitter to be used to stream the research. Defaults to None.	None
**kwargs	Any	Additional keyword arguments to be passed to the research method.	{}

Returns:

Name	Type	Description
LMOutput	LMOutput	The output from the language model.

PerplexityDeepResearcher(model_name='sonar-deep-research', api_key=None, prompt_builder=None, retry_config=None, streamable=True)

Bases: BaseDeepResearcher

A deep researcher that interacts with Perplexity deep research models.

This class extends the BaseDeepResearcher and interacts with Perplexity deep research models.

Examples:

deep_researcher = PerplexityDeepResearcher()
output = await deep_researcher.research(query=query, event_emitter=event_emitter)

Attributes:

Name	Type	Description
model_name	str	The name of the Perplexity deep research model to use.
client_params	dict[str, Any]	The parameters to be passed to the Perplexity client.
prompt_builder	PromptBuilder	The prompt builder to be used to format the inputs.
retry_config	RetryConfig	The retry configuration to be used to conduct the deep research.
streamable	bool	Whether the deep research will be streamed if an event emitter is provided.

Initializes a new instance of the PerplexityDeepResearcher class.

Parameters:

Name	Type	Description	Default
model_name	str	The name of the Perplexity deep research model to use. Defaults to "sonar-deep-research".	'sonar-deep-research'
api_key	str | None	The API key to use for the Perplexity deep research model. Defaults to None.	None
prompt_builder	PromptBuilder | None	The prompt builder to be used to format the inputs. Defaults to None, in which case a default prompt builder will be used.	None
retry_config	RetryConfig | None	The retry configuration to conduct the deep research. Defaults to a retry configuration with a timeout of DEFAULT_TIMEOUT.	None
streamable	bool	Whether the deep research will be streamed if an event emitter is provided. Defaults to True.	True

Raises:

Type	Description
ValueError	If the model name is not a deep research model.

research(query, history=None, extra_contents=None, hyperparameters=None, event_emitter=None, **kwargs) async

Conducts the deep research operation.

Parameters:

Name	Type	Description	Default
query	str	The query to research.	required
history	list[Message] | None	The conversation history to be considered in the research. Defaults to None.	None
extra_contents	list[MessageContent] | None	A list of extra contents to be included as inputs. Defaults to None.	None
hyperparameters	dict[str, Any] | None	The hyperparameters to be passed to the language model. Defaults to None.	None
event_emitter	EventEmitter | None	The event emitter to be used to stream the research. Defaults to None.	None
**kwargs	Any	Additional keyword arguments to be passed to the research method.	{}

Returns:

Name	Type	Description
LMOutput	LMOutput	The output from the language model.