Python API
The Python client calls public C++ driver methods over TCP using the generated drivers.json metadata.
Connect to a board
from koheron import connect
host = "192.168.1.100"
client = connect(host, name="fft")
connect() starts the named instrument if needed, opens a TCP connection and loads driver metadata. Additional arguments are passed to run_instrument().
Write a Python driver wrapper
@command() sends the command; the method body receives its result.
from koheron import command
class FFT:
def __init__(self, client):
self.client = client
@command()
def get_fft_size(self):
return self.client.recv_uint32()
@command()
def set_input_channel(self, channel):
pass
@command()
def read_psd(self):
return self.client.recv_array((4096,), dtype='float32')
Class and method names must match drivers.json, or be specified with @command(classname=..., funcname=...).
Application example
from koheron import connect
from fft import FFT
host = "192.168.1.100"
client = connect(host, name="fft")
fft = FFT(client)
print("FFT size:", fft.get_fft_size())
fft.set_input_channel(0)
psd = fft.read_psd()
print(psd[:10])
Upload and run an instrument
Run with the SDK's .venv/bin/python. Unlike make run, these calls return after the HTTP requests complete.
from koheron import upload_instrument, run_instrument
host = "192.168.1.100"
upload_instrument(host, "tmp/examples/alpha250/fft/fft.zip", run=True)
run=True starts the uploaded instrument. Restart a stored instrument with run_instrument(host, "fft", restart=True).
Query instrument status
from koheron import instrument_status
status = instrument_status("192.168.1.100")
print(status)
instruments lists stored names; live_instrument identifies the selected instrument.
Receive typed results
Receive helpers match the C++ return type:
uint32_t:recv_uint32()int32_t:recv_int32()float:recv_float()double:recv_double()bool:recv_bool()std::string:recv_string()- JSON string:
recv_json() std::array<T, N>:recv_array(shape, dtype)std::vector<T>:recv_vector(dtype)std::tuple<...>:recv_tuple(fmt)
The client checks return-type metadata where available and raises an error if the receive helper does not match the exported C++ return type.
Send arrays and vectors
Arguments support scalars, strings, std::array and std::vector types declared in the metadata. Use NumPy arrays with matching dtype for array and vector arguments.
