Quick start

Install the package, adapt your data once, and call an algorithm. There is no configuration file, no client to construct, and no API key — every function takes plain arrays and plain objects.

bash

npm install fintech-algorithms

Requires Node 22 or newer. The package is ESM-only and ships its own types.

Your first call#

Each algorithm has its own import path, and that path is identical to the URL of its article and of this page:

import { calculateEma } from "fintech-algorithms/technical-indicators/trend-smoothing/ema";

calculateEma([10, 13, 16, 19], 3);
// → [null, null, 13, 16]

Two things in that output are worth understanding before you go further, because they apply to almost every algorithm in the library.

`null` means "not yet defined"#

A 3-period average cannot exist until three observations have arrived. The first two positions are null — not 0, not omitted, not a partial average over one value. The returned array is always the same length as the input, so index i of the output always corresponds to index i of the input.

That matters the moment you plot it or join it to anything:

const ema = calculateEma(closes, 20);

// Safe: positions line up with your bars
bars.map((bar, i) => ({ time: bar.time, value: ema[i] }))
    .filter((point) => point.value !== null);

Dropping the nulls without keeping the index would silently shift your series backwards by the warm-up length. Every reference page states the warm-up count for its algorithm.

Bad input is a verdict, not an exception#

The validators and classifiers return a result per row instead of throwing, so a single bad tick cannot abort a whole batch:

import { calculate } from "fintech-algorithms/market-data-engineering/cleaning-and-validation/ohlc-consistency-validator";

const verdicts = calculate(bars);
const clean = bars.filter((_, i) => verdicts[i].status === "ok");

Pure arithmetic errors — a window of zero, an empty series — do throw, because those are programming mistakes rather than data problems.

Bring your own data#

The library ships no data provider: no HTTP client, no vendor SDK, no node:fs, and zero runtime dependencies. You write one small mapping function from your provider's payload to the shapes the library expects, and you own it.

import type { Bar } from "fintech-algorithms";

// Your provider → the library's Bar contract. One file, yours to keep.
const toBars = (payload: ProviderResponse): Bar[] =>
  payload.results.map((r) => ({
    timestamp: new Date(r.t).toISOString(),
    symbol: r.sym,
    open: r.o,
    high: r.h,
    low: r.l,
    close: r.c,
    volume: r.v,
  }));

When the vendor changes their API you edit that one file. The algorithms never move. See Wiring up a data provider for a complete worked adapter.

Finding the algorithm you need#

Three ways in:

Search — the box on the home page matches names, families and import paths.
Browse by domain — corporate actions, index construction, market breadth, and so on.
Derive the path — if you know the article URL, you know the import path.

If you write code with an agent, install the skill instead and let it do the looking up — it reads the same contracts and refuses to invent a path that does not exist: