TickNet data boundary

Language: English · 简体中文

Generic market-data ingestion, normalization, quality checks, versioning, and publication belong to quant-market-data-platform. The current ticknet package does not connect directly to that project's providers or published assets. Downstream research projects supply versioned data manifests and labels.

Ownership

quant-market-data-platform owns provider integrations, raw landing, canonical schemas, general-purpose data-quality checks, provenance, and published assets.

ticknet owns model-specific event-stream packing and loading, window construction, feature representations, labels, leakage checks, training splits, tensor materialization, training, replay, and evaluation.

quant-market-data-platform
  provider -> ingest -> raw -> standardize -> canonical -> quality/provenance -> published assets

downstream research project
  versioned inputs and labels -> ticknet datasets/windows -> train/evaluate

Cross-project handoff is managed by downstream projects. Reusable market-data cleaning belongs in quant-market-data-platform; transformations used only for model inputs belong in ticknet. This ownership boundary does not imply that this repository already implements a QMD adapter or a particular published-file contract.

Enforced boundary

The current implementation includes dataset, packing, and loading code under ticknet.eventstream. It does not contain the former documentation's canonical_adapter, nextday.snapshot_features, or nextday.snapshot_io modules. tests/test_microstructure_data_boundary.py checks that TickNet source and runtime dependencies do not directly import or depend on quant_market_data_platform, tushare, or rqdatac. A future schema-only distribution should be reviewed separately before adding it as a dependency.