xr_ai_tools#

Toolkit-independent native XR tools.

Submodules#

Classes#

`AsyncTool`	A validated asynchronous tool independent of any output transport.
`Tool`	A Pydantic-validated tool shared by agents, voice, and background triggers.
`ToolInvocationResult`	A model-visible result from one local tool invocation.
`ToolSet`	A native tool catalog with model-visible dispatch names.

Package Contents#

class xr_ai_tools.AsyncTool( name: str, description: str, request_model: type[RequestT], chunk_model: type[ChunkT], handler: collections.abc.Callable[[RequestT], collections.abc.AsyncIterator[ChunkT]], )#

A validated asynchronous tool independent of any output transport.

async stream( request: RequestT | collections.abc.Mapping[str, object], ) → collections.abc.AsyncIterator[ChunkT]#

Validate and yield chunks produced in an isolated Relay tool scope.

The producer task owns the scope and handler cleanup; chunks are yielded in the consumer’s context. Parent scope-local registrations are not transferred into the isolated context.

class xr_ai_tools.Tool( name: str, description: str, request_model: type[RequestT], result_model: type[pydantic.BaseModel] | None, handler: collections.abc.Callable[[RequestT], collections.abc.Awaitable[ResultT] | ResultT], *, return_direct: bool = False, render_result: collections.abc.Callable[[ResultT], str] | None = None, )#

A Pydantic-validated tool shared by agents, voice, and background triggers.

async execute(request: RequestT) → ResultT#: Run one validated request through the shared Relay tool lifecycle.

async invoke(arguments: str) → ToolInvocationResult#: Validate and run one model-supplied JSON argument payload.

class xr_ai_tools.ToolInvocationResult#

A model-visible result from one local tool invocation.

content: str#: Model-visible serialized result or validation error.

return_direct: bool#: Whether the agent should return the content without another model turn.

class xr_ai_tools.ToolSet( tools: collections.abc.Iterable[Tool[Any, Any]] | collections.abc.Mapping[str, Tool[Any, Any]], )#

A native tool catalog with model-visible dispatch names.

classmethod namespaced( namespaces: collections.abc.Mapping[str, collections.abc.Iterable[Tool[Any, Any]]], ) → ToolSet#: Build a catalog named <namespace>__<tool> for each group.

get(name: str) → Tool[Any, Any] | None#: Return the named tool when this catalog owns it.

items() → tuple[tuple[str, Tool[Any, Any]], Ellipsis]#: Return model-visible names paired with their underlying tools.