#!/usr/bin/env python3
"""Example: Load a Streaming Omni dataset config and inspect a row."""
from datasets import load_dataset

REPO = "skyzhou06/StreamingOmniDatasets"
CONFIG = "StreamingAudioInteraction"

print(f"Loading {REPO} config={CONFIG} train split (streaming=True)...")
try:
    ds = load_dataset(REPO, CONFIG, split="train", streaming=True)
    sample = next(iter(ds))
    print(f"\nSample keys: {list(sample.keys())}")
    print(f"id: {sample.get('id')}")
    print(f"dataset: {sample.get('dataset')}")
    print(f"task_family: {sample.get('task_family')}")
    print(f"duration_sec: {sample.get('duration_sec')}")
    print(f"modalities: {sample.get('modalities')}")

    timeline = sample.get("timeline", [])
    print(f"timeline chunks: {len(timeline)}")

    metadata = sample.get("metadata", {})
    print(f"model_mode: {metadata.get('model_mode')}")

    if timeline:
        t0 = timeline[0]
        print(f"\nFirst timeline chunk:")
        print(f"  t_start: {t0.get('t_start')}, t_end: {t0.get('t_end')}")
        print(f"  control: {t0.get('control')}")
        print(f"  asr_delta: {str(t0.get('asr_delta'))[:100]}")

except Exception as e:
    print(f"\nConfig-based loading failed: {e}")
    print("\nTry loading with data_files/parquet fallback:")
    print("  ds = load_dataset('skyzhou06/StreamingOmniDatasets',")
    print("      data_files='https://huggingface.co/datasets/skyzhou06/StreamingOmniDatasets/resolve/main/data/train-*.parquet',")
    print("      split='train', streaming=True)")