How wyze69-sys/quantara works
Quantara is archived research infrastructure for testing preregistered market-data features, not a trading or production-execution system. Its principal workflow acquires and verifies source data, builds canonical series and derived research tables, constructs validation folds, then trains candidates and evaluates them under frozen protocols. Supporting branches cover immutable publication and provenance, protocol-defined features and labels, and deterministic estimators. The README documents additional protocol versions and estimator families; their detailed wiring is not fully visible in the sampled source, so uncertain links are marked dashed or omitted.
Main entry points: research/pipeline.py Feature research table · publication.py Immutable publication · data_foundation/pipeline.py Data foundation pipeline · series/pipeline.py Series ingestion
Components 25
Data foundation
- Archive acquisition
src/quantara/data_foundation/acquisition.py - Descriptors and rights
src/quantara/data_foundation/descriptor.py - Canonical rows
src/quantara/data_foundation/canonical.py - Identity hashing
src/quantara/data_foundation/hashing.py - Canonical JSON
src/quantara/data_foundation/jcs.py - Quality approval
src/quantara/data_foundation/quality_approval.py - Immutable publication
src/quantara/publication.py - Data foundation pipeline
src/quantara/data_foundation/pipeline.py - Hash kernel
kernel/src/lib.rs
Market series
- Series ingestion
src/quantara/series/pipeline.py - Canonical series
src/quantara/series_canonical.py - Series parsing
src/quantara/series/parsing.py
Research tables
- Series derivation
src/quantara/derivation/pipeline.py - Feature research table
src/quantara/research/pipeline.py
Modeling and evaluation
- Walk-forward validation
src/quantara/validation/pipeline.py - Model training
src/quantara/training/pipeline.py - Deterministic estimators
src/quantara/estimator_c3.py - Logistic IRLS
src/quantara/training_metrics_logistic.py - Feature evaluation
src/quantara/evaluation/pipeline.py
Frozen protocols
- Protocol features
src/quantara/protocol_features.py - Protocol labels
src/quantara/protocol_labels.py - Protocol run
src/quantara/protocol_run.py - Protocol evaluation
src/quantara/protocol_evaluation_pipeline.py - V2 candidates
src/quantara/protocols/v2/dev_orchestrator.py
Other components
- Researcher
Connections 35
- Researcher → Data foundation pipeline: runs ingestion
- Data foundation pipeline → Archive acquisition: acquires archives
src/quantara/data_foundation/pipeline.py - Data foundation pipeline → Canonical rows: normalizes data
src/quantara/data_foundation/pipeline.py - Data foundation pipeline → Quality approval: checks quality
src/quantara/data_foundation/pipeline.py - Data foundation pipeline → Immutable publication: publishes artifacts
src/quantara/data_foundation/pipeline.py - Archive acquisition → Descriptors and rights: validates against
src/quantara/data_foundation/acquisition.py - Archive acquisition → Identity hashing: verifies bytes
src/quantara/data_foundation/acquisition.py - Canonical rows → Descriptors and rights: uses identity
src/quantara/data_foundation/canonical.py - Canonical rows → Identity hashing: defines content for
src/quantara/data_foundation/canonical.py - Identity hashing → Canonical JSON: canonicalizes with
src/quantara/data_foundation/hashing.py - Series ingestion → Archive acquisition: reuses acquisition
src/quantara/series/pipeline.py - Series ingestion → Series parsing: parses sources
src/quantara/series/pipeline.py - Series ingestion → Canonical series: persists series
src/quantara/series/pipeline.py - Series ingestion → Immutable publication: publishes artifacts
src/quantara/series/pipeline.py - Canonical series → Series parsing: uses parsed rows
src/quantara/series_canonical.py - Series derivation → Canonical rows: aggregates canonical data
src/quantara/derivation/pipeline.py - Series derivation → Immutable publication: publishes derived table
src/quantara/derivation/pipeline.py - Feature research table → Series derivation: authenticates parent
src/quantara/research/pipeline.py - Feature research table → Immutable publication: publishes research table
src/quantara/research/pipeline.py - Feature research table → Protocol features: computes feature table
- Protocol run → Protocol features: builds features
src/quantara/protocol_run.py - Protocol features → Protocol labels: uses labels
src/quantara/protocol_features.py - Protocol features → Canonical series: uses canonical series
src/quantara/protocol_features.py - Walk-forward validation → Feature research table: validates parent table
src/quantara/validation/pipeline.py - Walk-forward validation → Immutable publication: publishes folds
src/quantara/validation/pipeline.py - Model training → Feature research table: loads research data
src/quantara/training/pipeline.py - Model training → Walk-forward validation: loads folds
src/quantara/training/pipeline.py - Model training → Deterministic estimators: fits candidates
- Deterministic estimators → Logistic IRLS: uses logistic engine
- Feature evaluation → Feature research table: loads research data
src/quantara/evaluation/pipeline.py - Feature evaluation → Walk-forward validation: loads validation folds
src/quantara/evaluation/pipeline.py - Feature evaluation → Immutable publication: publishes evaluation
src/quantara/evaluation/pipeline.py - Protocol evaluation → Protocol run: evaluates run
src/quantara/protocol_evaluation_pipeline.py - V2 candidates → Deterministic estimators: selects candidates
- Identity hashing → Hash kernel: may use kernel