dragon.ai.agent.tools.registry.ToolRegistry

class ToolRegistry

Bases: object

Registry of BaseTool instances available to an agent.

Usage:

registry = ToolRegistry()
registry.register(MyTool())
tool = registry.get("my_tool")
schemas = registry.list_tools()   # OpenAI-compatible function schemas

__init__() → None

Methods

__init__()
get(name)	Return the tool registered under name.
has(name)	Return True if a tool with name is registered.
list_tools()	Return a list of OpenAI-compatible tool/function schemas.
register(tool)	Register a tool.
tool(fn)	Decorator that wraps fn as a FunctionTool and registers it.
tool_names()	Return sorted list of registered tool names.
unregister(name)	Remove a tool by name.

__init__() → None

register(tool: BaseTool | Callable ) → None

Register a tool.

tool can be a BaseTool instance or a plain callable (sync or async). Callables are automatically wrapped in FunctionTool.

Overwrites any existing tool with the same name.

unregister(name: str ) → None

Remove a tool by name. No-op if not present.

get(name: str ) → BaseTool

Return the tool registered under name.

Raises

KeyError

If no tool with that name is registered.

has(name: str ) → bool

Return True if a tool with name is registered.

list_tools() → list [dict [str , Any ]]

Return a list of OpenAI-compatible tool/function schemas.

These schemas are injected into the LLM system prompt so the model can decide which tools to call.

tool_names() → list [str ]

Return sorted list of registered tool names.

__len__() → int

tool(fn: Callable) → FunctionTool

Decorator that wraps fn as a FunctionTool and registers it.

Works with both sync and async functions:

registry = ToolRegistry()

@registry.tool
def calculate_magic_addition(a: int, b: int) -> int:
    """Add two numbers with a magic twist."""
    return a + b + 5

@registry.tool
async def fetch_data(url: str) -> dict:
    """Fetch data from a URL."""
    return await do_fetch(url)

Returns:

The created FunctionTool instance.