fix: enable tool calling for ChatMLX (fixes langchain-ai/langchain#31…

[Ujjwal-Bajpayee]

What does this PR do?

Adds support for tool/function calling in ChatMLX using MLXPipeline.

Why is this needed?

Fixes a bug where .tool_calls was always empty even when a model (e.g. Phi-4-reasoning-plus-MLX-4bit) supports tool calling.
Reference: langchain-ai/langchain#31726

Changes made:

• Injects tool schema into the prompt when tools are bound
• Parses .tool_calls from the model’s output (if JSON structured)
• Keeps the class structure and logic unchanged otherwise

[matiasdev30]

Hi still returning empty list [] in res.tool_calls

`from langchain_community.chat_models.mlx import ChatMLX
from langchain_community.llms import MLXPipeline
from langchain_core.tools import tool

@tool
def multiply(a: int, b: int) -> int:
"""Multiplies two integers."""
return a * b

chat = ChatMLX(
llm=MLXPipeline.from_model_id("Phi-4-reasoning-plus-MLX-4bit")
).bind_tools([multiply])

res = chat.invoke("What is 2 multiplied by 3?")
#print(res.content)
print(res.tool_calls)`

[]

[Ujjwal-Bajpayee]

@matiasdev30 Please check this and let me know

[Ujjwal-Bajpayee]

@matiasdev30