python(feat): add HDF5 client-side detect_config

[wei-qlu]

What was changed

Implement HDF5 detect_config client-side. This handles compound datasets by treating the first field as time and remaining fields as value channels, single-column datasets use a root time dataset if present, and channel metadata is read from HDF5 attributes (matching the existing backend implementation).

Verification

Unit tests and manual testing with a local script, verifying the full import process succeeded:

• Creating a config and importing
• Calling detect_config, making adjustments, and importing

Example import workflow

"""
Supported HDF5 layouts:
  1. Groups with separate datasets — time_dataset and value_dataset point to different paths
       /group_name/timestamps   (1D)
       /group_name/values       (1D)

  2. 2D datasets — same path for both, differentiated by column index
       /dataset_name            (N x 2, col 0 = time, col 1 = value)

  3. Compound datasets — same path for both, differentiated by field name
       /dataset_name            (compound dtype with named fields, e.g. ('Timestamp', 'Sensor Reading'))
       Use time_field/value_field to select the appropriate fields.
"""

import os
from dotenv import load_dotenv

from sift_client import SiftClient, SiftConnectionConfig
from sift_client.sift_types.channel import ChannelDataType
from sift_client.sift_types.data_import import (
    Hdf5ImportConfig,
    Hdf5DataColumn,
    TimeFormat,
)

load_dotenv()

HDF5_FILE = "all_types.h5"
ASSET_NAME = "hdf5_all_types_example"


def get_client() -> SiftClient:
    return SiftClient(
        connection_config=SiftConnectionConfig(
            api_key=os.getenv("SIFT_API_KEY"),
            grpc_url=os.getenv("GRPC_API_URL"),
            rest_url=os.getenv("REST_API_URL"),
            use_ssl=False,
        )
    )


def main():
    client = get_client()
    
    # alternatively, call detect_config and make adjustments as necessary
    config = Hdf5ImportConfig(
        asset_name=ASSET_NAME,
        run_name="all-types-manual-run",
        time_format=TimeFormat.ABSOLUTE_UNIX_NANOSECONDS,
        data=[
            # Layout 1: Groups — separate timestamps/values datasets
            Hdf5DataColumn(
                name="channel_bool",
                data_type=ChannelDataType.BOOL,
                time_dataset="channel_bool/timestamps",
                value_dataset="channel_bool/values",
            ),
            Hdf5DataColumn(
                name="channel_int32",
                data_type=ChannelDataType.INT_32,
                time_dataset="channel_int32/timestamps",
                value_dataset="channel_int32/values",
            ),
            Hdf5DataColumn(
                name="channel_int64",
                data_type=ChannelDataType.INT_64,
                time_dataset="channel_int64/timestamps",
                value_dataset="channel_int64/values",
            ),
            Hdf5DataColumn(
                name="channel_uint32",
                data_type=ChannelDataType.UINT_32,
                time_dataset="channel_uint32/timestamps",
                value_dataset="channel_uint32/values",
            ),
            Hdf5DataColumn(
                name="channel_uint64",
                data_type=ChannelDataType.UINT_64,
                time_dataset="channel_uint64/timestamps",
                value_dataset="channel_uint64/values",
            ),
            Hdf5DataColumn(
                name="channel_string",
                data_type=ChannelDataType.STRING,
                time_dataset="channel_string/timestamps",
                value_dataset="channel_string/values",
            ),
            Hdf5DataColumn(
                name="channel_enum",
                data_type=ChannelDataType.INT_64,
                time_dataset="channel_enum/timestamps",
                value_dataset="channel_enum/values",
            ),
            Hdf5DataColumn(
                name="channel_bit_field",
                data_type=ChannelDataType.UINT_64,
                time_dataset="channel_bit_field/timestamps",
                value_dataset="channel_bit_field/values",
            ),
            Hdf5DataColumn(
                name="channel_bytes",
                data_type=ChannelDataType.BYTES,
                time_dataset="channel_bytes/timestamps",
                value_dataset="channel_bytes/values",
            ),
            # Layout 2: 2D datasets — same path, differentiated by column index
            Hdf5DataColumn(
                name="channel_double",
                data_type=ChannelDataType.DOUBLE,
                time_dataset="channel_double",
                value_dataset="channel_double",
                time_index=0,
                value_index=1,
            ),
            Hdf5DataColumn(
                name="channel_float",
                data_type=ChannelDataType.DOUBLE,
                time_dataset="channel_float",
                value_dataset="channel_float",
                time_index=0,
                value_index=1,
            ),
        ],
    )

    job = client.data_import.import_from_path(
        file_path=HDF5_FILE,
        asset=ASSET_NAME,
        config=config,
        show_progress=True,
    )

    result = job.wait_until_complete()
    result.get_import_run()

if __name__ == "__main__":
    main()