CASOON Open Source
kestrel-chartkit is technical analysis as plain, testable Rust. Feed it price bars one at a time and it returns indicator values, the current market regime and a composite signal that explains itself. It computes, nothing more: you bring the bars, a renderer of your choice draws.
cargo add kestrel-chartkit --git https://github.com/casoon/kestrel-chartkit --tag v0.41.0
The same streaming calls behind a whole chart: EMA 20 and EMA 50, support and resistance zones from confirmed pivots, volume, RSI(14) and its latest signals. kestrel-chartkit computes every value when this site is built; chartlet draws them.
Synthetic 4-hour bars of a fictitious CASOON share with invented prices, no real quotes.
Work in progress
Every indicator consumes one bar at a time and answers from the bar its warmup ends. The same code serves a backtest over years of history and a live feed.
Each indicator runs against expected values derived outside the crate, and a contract fixes every bar: whether a value exists, every field it carries and every alert it raises. The <a href="/kestrel-chartkit/docs/reference/indicators/">indicator reference</a> lists all of them.
Regime and composite score follow a handful of fixed thresholds. A direction against the regime is vetoed, and every signal carries its subscores. Events alert on the bar they are confirmed, never earlier.
The market regime per bar, from classify_regime with ADX and ATR, on the fictitious CASOON share. kestrel-chartkit computes every value; chartlet draws them when this site is built.
examples/site_showcase.rsPure computation
Build an indicator by name and feed it bars. The quickstart in the documentation runs the full example from the terminal above.
build_checked.on_bar and read the output.let params = HashMap::from([("rsi_len".to_string(), 14.0)]);
let mut rsi = build_checked("rsi", ¶ms)?;
let bar = Bar::try_new(1700000000, 100.0, 105.0, 95.0, 104.0, 1000.0)?;
if let Some(output) = rsi.on_bar(&bar) {
println!("RSI: {:.2}", output.value);
}ApplicabilityReason has two new variants (PairPriceAdjustmentDiffers, PairSessionDiffers), so exhaustive match expressions over it need an update.ApplicabilityReason gains StructureContinuousSession and StructureMethod gains PriceGaps and PointAndFigure, so exhaustive match expressions over them need the new arms. method_rule_catalog returns 17 rules instead of 8.EqualLevelLiquidityConfig gains the public field list_min_touches, so a struct literal must set it or end in ..Default::default().check_pair_applicability, with pair_requirements and pair_rule_catalog, reports when the two series of a two-series indicator differ in price adjustment or session.elliott::rule_catalog() lists the conditions of the pivot sequence in the order they are tested, per structure the variants, rules and guidelines with their identifiers, sentences and origin, and the threshold constants, among them the new EXTENDED_WAVE_MIN_RATIO of identify_extended_wave. Details: Elliott rules.atr, tolerance (absolute), and every level with at least list_min_touches pivots as bsl_<i>_* / ssl_<i>_* with price, pivots, tests, status and the events of the bar (pivot, test, sweep, breakout, stop hunt, reclaim). Swept levels stay listed. Details: Levels and zones in the output.Not yet. The crate is pre-1.0 and under active development, so indicator names, parameters and outputs can still change fundamentally between versions. Pin a release tag and expect to adapt when you upgrade.
No. It returns plain indicator and signal data; drawing is left to a renderer such as chartlet for static charts, used throughout this site, or kestrel-loom for an interactive browser chart, still in alpha.
No. It's a rule-based summary of indicator readings from fixed thresholds, with a direction against the regime vetoed and every signal carrying its subscores — not a forecast, and nothing here is investment advice.
No. It contains mathematics only: it fetches no market data, connects to no broker or exchange, and holds no credentials, keys, accounts or personal data — you bring the bars.