ZMQ write() fails with NumPy values — file path converts types but ZMQ path doesn’t #246

Hi @pradeeban @mayureshkothare @Rahuljagwani ,
I think I found a runtime bug in the ZMQ path that’s easy to miss but breaks real workloads.

In concore.py , file‑based write() converts NumPy values to native Python types (convert_numpy_to_python()), but the ZMQ path sends the raw val directly into send_json_with_retry(). If val contains numpy.float64, numpy.int64, or arrays, JSON serialization fails and the message is never sent (you just get an “Unexpected error during ZMQ write” log).

This makes mixed studies or ZMQ‑only studies brittle whenever NumPy is involved (which is almost always in control loops).

Where

concore.py, write() function, ZMQ branch.

Steps to reproduce

Create a ZMQ‑based study or a simple script.
Call concore.write("port_name", "data", [np.float64(1.0), np.int64(2)]).
Observe the error log and missing message.

Expected

ZMQ writes should accept NumPy values the same way file‑based writes do.

Actual

ZMQ send_json fails to serialize NumPy types and the message is dropped.

Suggested fix

Apply convert_numpy_to_python() before send_json_with_retry() in the ZMQ path. This keeps ZMQ and file behavior consistent.

If this sounds reasonable, I’m happy to open a PR with a small regression test as well.