Packages | Limen

📄️ Data

Fetch raw market data, form optional bars, and hand experiment-ready splits to the rest of Limen.

Define experiments, run parameter search, and promote successful runs into reusable inference objects.

Package ready-to-run Single-File Decoders that plug directly into the Universal Experiment Loop.

Provide the lower-level technical indicator library that manifests and custom prep functions build on.

Build higher-level model inputs and target helpers on top of raw bars and indicator columns.

Hold lightweight transform helpers for target construction and data cleanup.

Fit train-only feature scaling and apply it consistently across validation and test data.

Score classifier and regressor outputs so experiments can compare permutations on consistent metrics.

Turn finished experiment runs into analyzable tables for prediction quality, confusion behavior, backtests, and parameter diagnostics.

Inference-time cohort surfaces for aggregating decoder outputs from completed Limen experiments.

Evaluate aligned predictions as trading outcomes through either a vectorized snapshot or a sequential simulation.

Provide the account ledger used by sequential backtests to track positions and balances over time.

Hold the smaller cross-cutting helpers that support experiments without belonging to one primary domain package.