ToolRegistry

relationalai.agent.cortex.tool

ToolRegistry(tools=None)

Registry for RAI tools accessible to a Cortex agent.

Provides a unified interface for tool discovery, verbalization, explanation, and querying over a RAI model.

Examples

>>> registry = ToolRegistry()
>>> registry.add(model=model, description="Jaffle shop data")

Methods

.add()

ToolRegistry.add(
    model: Optional[rai.Model] = None,
    description: Optional[str] = None,
    verbalizer: Optional[Verbalizer] = None,
    queries: Optional[Queries] = None,
    tool: Optional[Tool] = None,
) -> ToolRegistry

Add a tool to the registry.

This method supports two calling patterns:

1. Provide a Tool instance directly via the tool parameter.
2. Provide model + description (creates DefaultTool internally with optional verbalizer and queries).

Parameters:

• model

  (Model, default: None) - RAI Model instance (required if tool is None).

• description

  (str, default: None) - Human-readable tool description (required if tool is None).

• verbalizer

  (Verbalizer, default: None) - Custom verbalizer (only used if tool is None).

• queries

  (Queries, default: None) - Queries provider (only used if tool is None).

• tool

  (Tool, default: None) - Pre-constructed Tool instance. If provided, other parameters are ignored.

Returns:

• ToolRegistry - Self.

Examples:

>>> registry = ToolRegistry()
>>> registry.add(model=model, description="Jaffle shop")

.has_tools()

ToolRegistry.has_tools() -> bool

Check if at least one tool is registered.

Returns:

• bool - True if one or more tools are registered.

.has_capability()

ToolRegistry.has_capability(capability: Capability) -> bool

Check if any tool in the registry has the specified capability.

Parameters:

• capability

  (Capability) - The capability to check for.

Returns:

• bool - True if any registered tool has the capability.

.tool_has_capability()

ToolRegistry.tool_has_capability(tool_id: int, capability: Capability) -> bool

Check if a specific tool has the specified capability.

Parameters:

• tool_id

  (int) - The numeric tool index to check.

• capability

  (Capability) - The capability to check for.

Returns:

• bool - True if the tool exists and has the capability.

.discover()

ToolRegistry.discover() -> dict

Discover all registered tools and their metadata.

Returns:

• dict - Dictionary with "models" key containing a list of tool metadata including model_id, description, key_concepts, and query info.

.verbalize()

ToolRegistry.verbalize(model_id: int) -> str

Get detailed model structure for a specific tool.

Parameters:

• model_id

  (int) - The numeric tool index to verbalize.

Returns:

• str - Natural language description of the model structure.

.explain()

ToolRegistry.explain(model_id: int, concept: str) -> str

Get detailed explanation of a concept within a specific tool.

Parameters:

• model_id

  (int) - The numeric tool index containing the concept.

• concept

  (str) - The concept name to explain.

Returns:

• str - Natural language explanation of the concept.

.query()

ToolRegistry.query(model_id: int, query_dict: dict, limit: int = 1000) -> Any

Execute a query against a specific tool.

Parameters:

• model_id

  (int) - The numeric tool index to query.

• query_dict

  (dict) - Query parameters with "id" and "args" keys.

• limit

  (int, default: 1000) - Maximum number of rows to return. Default: 1000.

Returns:

• Any - Query results, typically as a pandas DataFrame.

Returned By

 agent > cortex > tool
└──  ToolRegistry
    └──  add