Tools

llm_function works from tool definitions that are converted into:

• available_functions
• available_callables

These are the structures passed into workflow_auto_assembler for planning and execution.

Where Tools Can Come From

Current tool-loading paths include:

• in-memory callables
• standalone .py files
• Python modules

The lower-level tool definition helpers live in llm_function_tools.

Why Tools Matter

The workflow planner can only assemble workflows from the tools it is given.

That means tool descriptions and schemas affect:

• whether a task is considered possible
• which workflow steps are chosen
• whether a saved workflow is still reusable later

Tool Sources

The current llm_function runtime supports tool-loading through source objects such as:

• InMemoryToolSource
• PythonFileToolSource
• PythonModuleToolSource

These let you keep tool definitions close to the use case that needs them, or organize them into reusable modules.