fix(agents): correctly extract tool arguments from AWS Bedrock responses

[NIK-TIGER-BILL]

Problem

When using AWS Bedrock models (e.g., us.amazon.nova-pro-v1:0), all tool calls receive empty arguments {} regardless of what the LLM provides.

The bug is in crew_agent_executor.py:

# BEFORE (broken)
func_args = func_info.get("arguments", "{}") or tool_call.get("input", {})

func_info.get("arguments", "{}") returns the string "{}" (truthy!) when the key is absent. This means the or operator never evaluates tool_call.get("input", {}), so Bedrock's input field is always ignored.

Fix

Use .get() without a default to distinguish a missing key from an empty string, then fall back to the Bedrock input field only when the OpenAI arguments field is absent or empty:

# AFTER (fixed)
_raw_args = func_info.get("arguments")
func_args = (_raw_args if _raw_args not in (None, "", "{}") else None) or tool_call.get("input", {})

Fixes #4748

---

Note

Medium Risk
Touches core native tool-call parsing in CrewAgentExecutor, which can change what arguments are passed into tools across providers. Small, targeted logic change but impacts tool execution correctness.

Overview
Fixes native tool-call parsing in CrewAgentExecutor._parse_native_tool_call so missing OpenAI function.arguments no longer defaults to the truthy string "{}".

When arguments is absent/empty, the executor now correctly falls back to Bedrock-style tool call input, preventing tools from being invoked with empty {} arguments on Bedrock responses.

^{Written by Cursor Bugbot for commit 5c72dab. This will update automatically on new commits. Configure here.}