Workflow Cache

llm_function inherits workflow reuse behavior from workflow_auto_assembler.

That means workflow reuse can involve:

• in-memory cache
• filesystem storage
• workflow replanning when stale saved workflows no longer match the current toolset

Runtime Interaction

When you use LlmFunctionRuntime, repeated decorated calls reuse the same initialized WAA instance.

That also means they reuse the same in-memory workflow cache held by WAA.

More Detail

For deeper cache behavior, see the WAA docs:

• Workflow Cache
• Core Ideas