ta-crypto

Technical analysis indicators for crypto markets in Node.js.

Install

npm i ta-crypto

Quick Start

import { sma, rsi, macd, bbands, atr, vwapSession, toOHLCV } from "ta-crypto";

const candles = [
  { open: 100, high: 102, low: 99, close: 101, volume: 10, time: 1 },
  { open: 101, high: 103, low: 100, close: 102, volume: 12, time: 2 }
];

const { high, low, close, volume } = toOHLCV(candles, 0);
const s = sma(close, 14);
const r = rsi(close, 14);
const m = macd(close);
const b = bbands(close, 20, 2);
const a = atr(high, low, close, 14);
const vs = vwapSession(high, low, close, volume, [1, 1]);

Stateful API (streaming)

import { createSMA, createEMA, createRSI, createVWAPSession } from "ta-crypto";

const sma14 = createSMA(14);
const ema14 = createEMA(14);
const rsi14 = createRSI(14);
const nextSma = sma14.next(101.25);
const nextEma = ema14.next(101.25);
const nextRsi = rsi14.next(101.25);

const vwap = createVWAPSession();
const nextVwap = vwap.next({
  high: 102,
  low: 99,
  close: 101,
  volume: 10,
  sessionId: "2026-02-10-asia"
});

Websocket-like streaming loop:

import { createEMA, createRSI } from "ta-crypto";

const ema21 = createEMA(21);
const rsi14 = createRSI(14);

ws.on("message", (tick) => {
  const price = Number(tick.last);
  const e = ema21.next(price);
  const r = rsi14.next(price);
  if (e !== null && r !== null) {
    // strategy signal path
  }
});

Examples

Real-world entry points live in examples/:

• examples/rsi-strategy.ts
• examples/vwap-session.ts
• examples/funding-arbitrage.ts

If you're using ta-crypto in production or experiments, feel free to add your example here.

Compatibility

ta-crypto now ships golden tests (test/fixtures/golden.json) to lock behavior across releases.

Classic indicators:

| Indicator | Reference | Tolerance | | --- | --- | --- | | SMA, EMA, RSI | Golden baseline | 1e-10 | | MACD, BBANDS | Golden baseline | 1e-10 | | ATR, ADX | Golden baseline | 1e-10 |

Crypto indicators:

| Indicator | Reference | Tolerance | | --- | --- | --- | | VWAP Session | Golden baseline | 1e-10 | | Stateful VWAP Session parity | Batch VWAP Session | 1e-10 |

Streaming parity:

| Indicator | Reference | Tolerance | | --- | --- | --- | | Stateful RSI(14) | Batch RSI(14) | 1e-10 |

External parity:

Policy source: scripts/compat-policy.json

| Indicator | Burn-in | Tolerance | Blocking refs | Non-blocking refs | | --- | --- | --- | --- | --- | | SMA(14) | 14 | 1e-10 | TA-Lib, technicalindicators | pandas-ta | | EMA(14) | 14 | 1e-10 | TA-Lib, technicalindicators | pandas-ta | | RSI(14) | 28 | 5e-2 | TA-Lib, technicalindicators | pandas-ta | | MACD(12,26,9) | 80 | 2e-2 | TA-Lib, technicalindicators | pandas-ta | | BBANDS(20,2) | 20 | 1e-10 | TA-Lib, technicalindicators | pandas-ta | | ATR(14) | 56 | 1.5e-1 | TA-Lib, technicalindicators | pandas-ta | | ADX/+DI/-DI(14) | 90 | 1.5 | TA-Lib, technicalindicators | pandas-ta |

Warmup and alignment rules:

• References are left-padded to full series length before comparison.
• Comparison starts at indicator-specific burn-in and uses only overlapping non-null points.
• TA-Lib and technicalindicators are release-blocking checks in CI.
• pandas-ta is telemetry-only (warns when divergent or unavailable by environment).

Generate/review vectors:

npm run generate:golden
npm run test:golden
npm run generate:compat
npm run test:compat:technicalindicators
# python deps: pip install -r scripts/requirements-compat.txt
npm run test:compat:python

Note:

• CI uses Linux + Python 3.12 for full TA-Lib/pandas-ta coverage.
• On Windows, pandas-ta may be unavailable depending on upstream wheels; the script reports explicit skip in that case.
• TA-Lib and technicalindicators are strict release gates; pandas-ta runs as compatibility telemetry and reports warnings when it diverges by environment.

Crypto Playbooks

Session VWAP reset:

• Provide one sessionId per candle.
• VWAP resets exactly when sessionId changes.
• Use exchange session boundaries (UTC day, funding window, or custom market session).

Funding APR:

• fundingRateAPR(values, periodsPerYear) annualizes periodic funding.
• periodsPerYear presets:
• 3/day funding: 1095
• 1/hour funding: 8760
• 1/8h funding: 1095

Volatility regime:

• volatilityRegime(values, length, periodsPerYear, lowZ, highZ) returns -1, 0, 1.
• Default thresholds: lowZ = -0.5, highZ = 0.5.
• Calibration approach: increase absolute thresholds for noisier low-timeframe pairs, reduce for smoother higher timeframes.

Indicators

Overlap:

• sma, ema, rma, hl2, hlc3, ohlc4, vwap, bbands

Momentum:

• rsi, macd, stoch

Volatility:

• trueRange, atr, natr, realizedVolatility

Performance:

• logReturn, percentReturn

Volume:

• obv, mfi

Trend:

• adx

Crypto:

• vwapSession, fundingRateCumulative, fundingRateAPR, volatilityRegime, signedVolume, volumeDelta, orderflowImbalance

Hero Features (crypto edge)

1. Session-aware VWAP (vwapSession, createVWAPSession)
2. Funding analytics (fundingRateCumulative, fundingRateAPR)
3. Volatility regime labeling (volatilityRegime)
4. Orderflow proxies (signedVolume, volumeDelta, orderflowImbalance)
5. Streaming/stateful indicators for low-allocation pipelines

Limitations:

• Orderflow proxies infer pressure from candle direction and volume; they are not a replacement for L2/L3 order book data.
• Different libraries use different warmup conventions; comparisons use overlapping non-null windows.

Typing and Inputs

ta-crypto public APIs accept two input styles:

• Primitive arrays (number[]) for low-level control.
• Candle objects with long keys (open/high/low/close/volume/time) or aliases (o/h/l/c/v/t).

Single-series APIs (for example sma, ema, rsi, macd, bbands) accept:

import { rsi } from "ta-crypto";

const close = [101, 102, 103, 104];
const candles = [{ o: 100, h: 102, l: 99, c: 101, v: 10, t: 1 }];

rsi(close, 14);
rsi(candles, 14); // uses candle close/c

OHLC/OHLCV APIs (for example vwap, stoch, atr, natr, mfi, adx) accept:

import { atr, vwap } from "ta-crypto";

atr([102, 103], [99, 100], [101, 102], 14);
atr([{ open: 100, high: 102, low: 99, close: 101, volume: 10 }], 14);

vwap([102, 103], [99, 100], [101, 102], [10, 12]);
vwap([{ o: 100, h: 102, l: 99, c: 101, v: 10 }]);

Helpers:

import { pluckClose, toOHLCV } from "ta-crypto";

const close = pluckClose(candles);
const ohlcv = toOHLCV(candles, 0);
const ohlcvFromArrays = toOHLCV({ o: [1, 2], h: [3, 4], l: [0, 1], c: [2, 3], v: [5, 6] });

Validation:

• Multi-series APIs enforce equal lengths.
• Numeric inputs are validated as finite numbers with index-aware messages when possible.

Module Imports

import { sma } from "ta-crypto/indicators";
import { vwapSession } from "ta-crypto/crypto";
import { toOHLCV } from "ta-crypto/candles";
import { createSMA, createEMA, createRSI } from "ta-crypto/stateful";

Bench (internal baseline, 10k candles)

sma(14): 0.619 ms/run
ema(14): 0.549 ms/run
rsi(14): 0.647 ms/run
macd:    2.887 ms/run
bbands:  2.159 ms/run
atr(14): 1.356 ms/run
adx(14): 5.371 ms/run

Run locally:

npm run bench

Contributing

ta-crypto is open to contributors. If you use it, improve it. If you miss something, build it. If you're learning, this is a great codebase to practice financial math.

Start with CONTRIBUTING.md and ROADMAP.md. If you want to contribute but don't know where to start, pick an indicator from the roadmap and add tests comparing with TA-Lib.

Release

Publications are gated by GitHub Actions:

• CI runs build/tests + compatibility checks.
• Release workflows run tests/compat again before publish.

Recommended with GitHub CLI:

gh auth login
gh workflow run ci.yml
gh run list --workflow ci.yml --limit 1

npm run changelog
npm run version:patch
npm run release

License

MIT