srilekhaejnavarjala/yottaflex-chronos-ml-feature-pipeline

How srilekhaejnavarjala/yottaflex-chronos-ml-feature-pipeline works

Chronos is a local batch feature-engineering demonstration, not a model-training or deployed serving system. Its main workflow reads customer, transaction, and prediction-request CSVs; validates and quarantines rows; builds point-in-time features and separate future labels; publishes results to SQLite; then reconciles and exports run artifacts. The Streamlit dashboard exposes execution and result inspection, while deterministic data generation and a point-in-time demo are separate capabilities. Relationships follow the README and the sampled orchestrator; unsampled implementation details are not inferred.

Main entry points: pipeline.py Pipeline orchestrator · features.py Feature builder · publish.py SQLite store · ingest.py CSV ingestion

Components 17

User entry points

Streamlit dashboard app.py

Batch pipeline

Pipeline orchestrator src/pipeline.py
CSV ingestion src/ingest.py
Validation and quarantine src/validate.py
Feature builder src/features.py
Label builder src/labels.py
Run metrics src/metrics.py

Other components

Operator
Input CSVs

Connections 31

Operator → Streamlit dashboard: opens README.md
Operator → Pipeline orchestrator: runs README.md
Streamlit dashboard → Pipeline orchestrator: runs pipeline README.md
Input CSVs → CSV ingestion: provides CSVs README.md
Pipeline orchestrator → CSV ingestion: invokes src/pipeline.py
Pipeline orchestrator → Validation and quarantine: invokes src/pipeline.py
Pipeline orchestrator → Feature builder: invokes src/pipeline.py
Pipeline orchestrator → Label builder: invokes src/pipeline.py
Pipeline orchestrator → SQLite store: publishes through src/pipeline.py
Pipeline orchestrator → Reconciliation: runs checks src/pipeline.py
Pipeline orchestrator → Run metrics: records timings README.md
CSV ingestion → Validation and quarantine: passes raw rows README.md
Validation and quarantine → Feature builder: passes valid rows README.md
Validation and quarantine → SQLite store: quarantines rejects README.md
Feature builder → SQLite store: publishes features README.md
Label builder → SQLite store: publishes labels README.md
SQLite store → Reconciliation: checks stored data README.md
Pipeline orchestrator → CSV exports and reports: exports results README.md
SQLite store → CSV exports and reports: exports tables README.md
Synthetic data generator → Input CSVs: generates data README.md
Operator → Synthetic data generator: runs README.md
Operator → Point-in-time demo: runs README.md
CSV ingestion → Data contracts: uses contracts src/ingest.py
Validation and quarantine → Data contracts: uses contracts src/validate.py
Feature builder → Data contracts: uses contracts src/features.py
Feature builder → Feature configuration: reads settings src/features.py
Label builder → Feature configuration: reads settings src/labels.py
CSV ingestion → UTC time utilities: parses timestamps src/ingest.py
Feature builder → UTC time utilities: converts timestamps src/features.py
Label builder → UTC time utilities: converts timestamps src/labels.py
Reconciliation → Feature builder: audits parity src/reconcile.py

Drawn by GitDiagram on November 11, 2026 from the repository's file tree, README and source files. Read as Markdown ·