fix: use resource.name for mockable in process_tool

[mjnovice]

Summary

Fixes tool simulation (uipath debug) failing in two ways:

Bug 1: @mockable name mismatch

process_tool.py and escalation_tool.py used name=tool_name.lower() (e.g. api_workflow) instead of name=resource.name (e.g. API Workflow). This caused LLMMocker to raise UiPathNoMockFoundError because the name didn't match simulation.json's toolsToSimulate[].name.

Fallback behavior: When UiPathNoMockFoundError is raised, the @mockable decorator silently falls through to the real tool function (mockable.py:mocked_response_decorator). This means tools configured for simulation would execute against the actual UiPath platform instead, failing with HTTP 400 errors in
debug/eval contexts.

Bug 2: Empty kwargs in @mockable-decorated nested functions

Several tools define a nested function decorated with @mockable that takes no arguments (e.g. async def invoke_process():), even though the outer function receives **kwargs. The @mockable decorator captures function arguments to build the currentToolInput field in the mock LLM prompt. With no arguments,
the mock LLM sees {"args": [], "kwargs": {}} and returns null results, causing the agent to loop until max iterations.

Files changed

File	Name fix	Kwargs fix
process_tool.py	resource.name	invoke_process(**_tool_kwargs)
escalation_tool.py	resource.name	escalate(**_tool_kwargs)
deeprag_tool.py	already correct	invoke_deeprag(**_tool_kwargs)
batch_transform_tool.py	already correct	invoke_batch_transform(**_tool_kwargs)

No changes needed: integration_tool.py, context_tool.py, analyze_files_tool.py, mcp_tool.py, ixp_escalation_tool.py — these either decorate the top-level function directly (kwargs flow through naturally) or have no user-provided inputs.

Test plan

• Run uipath debug on an agent with a process tool configured for simulation — verify mock interception works and returns non-null results
• Run uipath debug on an agent with an escalation tool configured for simulation
• Verify non-simulated tools still execute normally (no regression)
• Run uipath eval to confirm evaluation pipeline works end-to-end

RCA

PR #546 introduced Bug 1 (name mismatch) for process_tool.py only.

The diff shows it moved the @mockable decorator from the outer process_tool_fn level to a new nested invoke_process() function, and in doing so made two changes:

1. Changed name=resource.name → name=tool_name.lower() (Bug 1)
2. The new nested invoke_process() was defined with no arguments and called with no arguments (Bug 2)

🤖 Generated with Claude Code