Skip to main content

limen.indicators

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

Canonical docs

What this package owns

Owns the public indicator function library, including moving averages, oscillators, volatility measures, price transforms, and candlestick-pattern helpers. Does not own higher-level feature engineering, target creation, or model training.

Key entry points

Entry pointUse it whenNotes
limen.indicators.* exportsYou want indicators in a manifest or custom prep pipelineThe package root re-exports the public indicator surface
sma, ema, wilder_rsi, atrYou need common directional and volatility baselinesFrequently used in foundational SFDs
bbands, bollinger_bands, bollinger_positionYou want band-based regime or position signalsUseful before higher-level feature construction
macd, ppo, roc, stoch, stochrsiYou need momentum and oscillator familiesMost emit multiple columns or parameterized output names
Candlestick helpers like cdldojiYou want pattern flags aligned with TA-Lib semanticsExported at the package root alongside the rest of the library

Adjacent modules

  • limen.features typically consumes indicator columns and turns them into more opinionated model inputs.
  • limen.experiment.Manifest wires indicators into the prep pipeline through .add_indicator(...).
  • limen.data supplies the OHLCV frames that most indicators expect.

Quick orientation

indicators/
├── ma, sma, ema, wma, tema, trima, t3 ... # Moving averages and trend baselines
├── rsi, wilder_rsi, cmo, willr, ultosc ... # Momentum and oscillator families
├── atr, natr, trange, stddev, var ... # Volatility and range measures
├── bbands, bollinger_bands, midpoint ... # Band and price-position helpers
├── macd, macdfix, macdext, ppo ... # Multi-column momentum helpers
└── cdl*.py # Candlestick pattern detections

Things to know

  • Most indicator functions accept a pl.LazyFrame and return that frame with one or more appended columns.
  • Leading null rows are expected for rolling calculations. Manifest-driven prep drops them after the feature and indicator stage.
  • The public reference should be treated as the canonical list of exported helpers and output-column conventions.
  • Many indicators aim for TA-Lib parity, so naming and behavior are intentionally close to that surface where applicable.