Create or update an agent

async AsyncCogniteClient.agents.upsert(

agents: AgentUpsert | Sequence[AgentUpsert],

) → Agent | AgentList

Create or update (upsert) one or more agents.

Parameters:

agents (AgentUpsert | Sequence[AgentUpsert]) – Agent or list of agents to create or update.

Returns:

The created or updated agent(s).

Return type:

Agent | AgentList

Examples

Create a new agent with a query knowledge graph tool to find assets:

>>> from cognite.client import CogniteClient
>>> from cognite.client.data_classes.agents import (
...     AgentUpsert,
...     QueryKnowledgeGraphAgentToolUpsert,
...     QueryKnowledgeGraphAgentToolConfiguration,
...     DataModelInfo,
... )
>>> client = CogniteClient()
>>> find_assets_tool = QueryKnowledgeGraphAgentToolUpsert(
...     name="find assets",
...     description="Use this tool to find assets",
...     configuration=QueryKnowledgeGraphAgentToolConfiguration(
...         data_models=[
...             DataModelInfo(
...                 space="cdf_idm",
...                 external_id="CogniteProcessIndustries",
...                 version="v1",
...                 view_external_ids=["CogniteAsset"],
...             )
...         ]
...     ),
... )
>>> agent = AgentUpsert(
...     external_id="my_agent",
...     name="My Agent",
...     labels=["published"],
...     tools=[find_assets_tool],
... )
>>> client.agents.upsert(agents=[agent])

Create an agent with multiple different tools:

>>> from cognite.client.data_classes.agents import (
...     AgentUpsert,
...     QueryKnowledgeGraphAgentToolUpsert,
...     QueryKnowledgeGraphAgentToolConfiguration,
...     DataModelInfo,
...     SummarizeDocumentAgentToolUpsert,
...     AskDocumentAgentToolUpsert,
...     QueryTimeSeriesDatapointsAgentToolUpsert,
... )
>>> find_assets_tool = QueryKnowledgeGraphAgentToolUpsert(
...     name="find assets",
...     description="Use this tool to query the knowledge graph for assets",
...     configuration=QueryKnowledgeGraphAgentToolConfiguration(
...         data_models=[
...             DataModelInfo(
...                 space="cdf_idm",
...                 external_id="CogniteProcessIndustries",
...                 version="v1",
...                 view_external_ids=["CogniteAsset"],
...             )
...         ]
...     ),
... )
>>> find_files_tool = QueryKnowledgeGraphAgentToolUpsert(
...     name="find files",
...     description="Use this tool to query the knowledge graph for files",
...     configuration=QueryKnowledgeGraphAgentToolConfiguration(
...         data_models=[
...             DataModelInfo(
...                 space="cdf_idm",
...                 external_id="CogniteProcessIndustries",
...                 version="v1",
...                 view_external_ids=["CogniteFile"],
...             )
...         ]
...     ),
... )
>>> find_time_series_tool = QueryKnowledgeGraphAgentToolUpsert(
...     name="find time series",
...     description="Use this tool to query the knowledge graph for time series",
...     configuration=QueryKnowledgeGraphAgentToolConfiguration(
...         data_models=[
...             DataModelInfo(
...                 space="cdf_idm",
...                 external_id="CogniteProcessIndustries",
...                 version="v1",
...                 view_external_ids=["CogniteTimeSeries"],
...             )
...         ]
...     ),
... )
>>> summarize_tool = SummarizeDocumentAgentToolUpsert(
...     name="summarize document",
...     description="Use this tool to get a summary of a document",
... )
>>> ask_doc_tool = AskDocumentAgentToolUpsert(
...     name="ask document",
...     description="Use this tool to ask questions about specific documents",
... )
>>> ts_tool = QueryTimeSeriesDatapointsAgentToolUpsert(
...     name="query time series",
...     description="Use this tool to query time series data points",
... )
>>> agent = AgentUpsert(
...     external_id="my_agent",
...     name="My agent",
...     description="An agent with many tools",
...     instructions="You are a helpful assistant that can query knowledge graphs, summarize documents, answer questions about documents, and query time series data points.",
...     labels=["published"],
...     tools=[
...         find_assets_tool,
...         find_files_tool,
...         find_time_series_tool,
...         summarize_tool,
...         ask_doc_tool,
...         ts_tool,
...     ],
... )
>>> client.agents.upsert(agents=[agent])