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 point | Use it when | Notes |
|---|---|---|
limen.indicators.* exports | You want indicators in a manifest or custom prep pipeline | The package root re-exports the public indicator surface |
sma, ema, wilder_rsi, atr | You need common directional and volatility baselines | Frequently used in foundational SFDs |
bbands, bollinger_bands, bollinger_position | You want band-based regime or position signals | Useful before higher-level feature construction |
macd, ppo, roc, stoch, stochrsi | You need momentum and oscillator families | Most emit multiple columns or parameterized output names |
Candlestick helpers like cdldoji | You want pattern flags aligned with TA-Lib semantics | Exported at the package root alongside the rest of the library |
Adjacent modules
limen.featurestypically consumes indicator columns and turns them into more opinionated model inputs.limen.experiment.Manifestwires indicators into the prep pipeline through.add_indicator(...).limen.datasupplies 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.LazyFrameand return that frame with one or more appended columns. - Leading
nullrows 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.