Grencape/semantaai-crypto_assets

# SemantaAI Dataset Standard

This repository follows a 2-layer standard with no versioned folder names.

## Layout
- `raw/`: canonical OHLCV history (Binance spot, 5m, one file per asset).
- `gold/`: enriched training-ready layer (`raw` + feature engineering + labels + targets).

## Raw
- Contains original OHLCV parquet files and per-asset metadata.
- Main index files:
- `raw/DATASET_INDEX.json`
- `raw/DATASET_INDEX.csv`

## Gold
- Contains the production feature/label layer used for model training.
- Includes deviation features around moving averages (`SMA/EMA/DEMA/TEMA`) and trend/mean-reversion confidence signals.
- Main docs and indexes:
- `gold/README.md`
- `gold/METRICS.md`
- `gold/DATASET_INDEX.json`
- `gold/DATASET_INDEX_FILES.json`

## Notes
- Legacy folders `labeled_v1/` and `labeled_v2/` were consolidated into `gold/`.
- Standard policy for all future SemantaAI datasets:
- `DATA_STANDARD.md`