# Phase 1 Report

Generated at UTC: 2026-06-16T13:31:25.599948+00:00

## Exact Row Counts

| Dataset | Train | Eval | High quality | Average duration sec |
| --- | ---: | ---: | ---: | ---: |
| StreamingAudioInteraction | 950 | 50 | 1000 | 4.18 |
| StreamingASRTranslation | 950 | 50 | 1000 | 5.448 |
| SpeechOutputAlignment | 950 | 50 | 1000 | 5.346 |

## Source Distribution

### StreamingAudioInteraction
- local_text_dialogues_fallback: 1000

### StreamingASRTranslation
- local_speech_transcripts_fallback: 1000

### SpeechOutputAlignment
- StreamingAudioInteraction: 1000

## Control Label Distribution

### StreamingAudioInteraction
- response: 1000
- silent: 4100

### StreamingASRTranslation
- listen: 4600
- response: 1300

### SpeechOutputAlignment
- speak: 6000

## Failed Or Rejected Samples

- No failed samples were written to final high_quality outputs.

## Missing Dependencies Or Sources

- WhisperX not detected; deterministic timestamp fallback used: reported by 3 dataset(s)
- CosyVoice command/repo not configured; fallback tone TTS used: reported by 3 dataset(s)

## Environment

- python: `3.13.9`
- platform: `macOS-26.5.1-arm64-arm-64bit-Mach-O`
- pandas: `ok`
- pyarrow: `ok`
- yaml: `ok`
- numpy: `ok`
- soundfile: `ok`
- datasets: `ok`
- whisperx: `fallback_alignment`
- cosyvoice: `fallback_tone_tts`
- ffmpeg: `/opt/anaconda3/bin/ffmpeg`

## Source Preparation

- phase: `phase1`
- dialogue_fallback_path: `/Users/skyzhou/Desktop/StreamingDatasets/data/raw/text_dialogues/fallback_dialogues.jsonl`
- dialogue_rows: `1500`
- dialogue_created_or_rewritten: `True`
- speech_fallback_path: `/Users/skyzhou/Desktop/StreamingDatasets/data/raw/local_speech_transcripts/fallback_speech_transcripts.jsonl`
- speech_rows: `1500`
- speech_created_or_rewritten: `True`
- note: `Fallback sources are local synthetic text/transcript seeds for pipeline smoke tests and empty-repo operation.`

## Validation

- status: `pass`
- StreamingAudioInteraction: `pass` with 0 errors
- StreamingASRTranslation: `pass` with 0 errors
- SpeechOutputAlignment: `pass` with 0 errors

## Commands Executed

- 2026-06-16T13:21:38.265459+00:00: `scripts/00_check_env.py`
- 2026-06-16T13:21:42.366217+00:00: `scripts/01_prepare_sources.py --phase phase1`
- 2026-06-16T13:21:54.375310+00:00: `scripts/04_build_streaming_audio_interaction.py --target-rows 10 --force`
- 2026-06-16T13:21:58.893165+00:00: `scripts/05_build_streaming_asr_translation.py --target-rows 10 --force`
- 2026-06-16T13:22:16.167739+00:00: `scripts/04_build_streaming_audio_interaction.py --target-rows 10 --force`
- 2026-06-16T13:22:22.049161+00:00: `scripts/05_build_streaming_asr_translation.py --target-rows 10 --force`
- 2026-06-16T13:22:27.575785+00:00: `scripts/06_build_speech_output_alignment.py --target-rows 10 --force`
- 2026-06-16T13:22:31.807805+00:00: `scripts/07_validate_datasets.py`
- 2026-06-16T13:22:36.387527+00:00: `scripts/08_make_preview.py --limit 20`
- 2026-06-16T13:22:51.732206+00:00: `scripts/04_build_streaming_audio_interaction.py --target-rows 1000`
- 2026-06-16T13:22:57.870130+00:00: `scripts/05_build_streaming_asr_translation.py --target-rows 1000`
- 2026-06-16T13:23:10.715232+00:00: `scripts/06_build_speech_output_alignment.py --target-rows 1000`
- 2026-06-16T13:23:16.940626+00:00: `scripts/07_validate_datasets.py`
- 2026-06-16T13:23:25.849259+00:00: `scripts/08_make_preview.py --limit 20`
- 2026-06-16T13:24:34.682757+00:00: `scripts/07_validate_datasets.py`
- 2026-06-16T13:24:41.520226+00:00: `scripts/08_make_preview.py --limit 20`
- 2026-06-16T13:31:22.464889+00:00: `scripts/07_validate_datasets.py`
- 2026-06-16T13:31:24.981472+00:00: `scripts/08_make_preview.py --limit 20`

## Next Steps For Phase 2

- Replace fallback tone audio with real CosyVoice output by setting `tools.cosyvoice.command_template` or repository paths.
- Run external WhisperX batch alignment through `scripts/02_run_whisperx_align.py` when the deployed path is known.
- Add licensed non-benchmark speech and dialogue sources under `data/raw/` and rebuild with `--force`.
- Add video and OCR fields only after a source with reliable visual evidence is available.