Indicators

The indicator node runs any of 116 technical indicators from the tradeboard.ta library over a symbol's history, server-side.

Basic use

{ "id": "r", "type": "indicator", "position": { "x": 0, "y": 100 },
  "data": {
    "symbol": "RELIANCE", "exchange": "NSE",
    "interval": "D", "source": "api",
    "indicatorName": "rsi",
    "params": "{\"period\": 14}",
    "lookbackBars": 100,
    "tailBars": 5,
    "offsetBars": 0,
    "outputVariable": "rsi"
  } }

Field	Meaning
symbol, exchange	What to compute over. Not needed in nested mode.
interval	Timeframe, free text (1m, 5m, 15m, 1h, D, …). Use the intervals node to list what your broker supports.
source	The data source, api (broker) or db (Historify, resampled locally). Not a price field.
indicatorName	Function name, lowercase (rsi, sma, supertrend, macd, …)
params	JSON object of the indicator's own arguments, as a string
lookbackBars	How much history to compute over (capped at 200)
tailBars	Length of the returned series array (also capped at 200)
offsetBars	Which bar at_offset should read (0 = same bar as latest, max 199)
sourceSeries	Optional: compute over another series instead of fetching
sourceField	Optional: which field of a row to read from each sourceSeries row, the price field (close, high, low) or a specific out1

The node fetches lookbackBars plus a fixed 50-bar warm-up so the indicator has settled by the time the requested window starts, and the sum is still capped at 200. That is why {{rsi.bars_used}} can be larger than the lookbackBars you asked for.

source is not sourceField

They are two different things and the names invite the mix-up:

• source selects where the OHLCV comes from: api calls the broker, db reads Historify and resamples locally (2m/3m/25m/2h out of stored 1m, W/M/Q/Y out of D). Nothing else is valid.
• sourceField selects which field of a row to read, and only applies in nested mode, where each sourceSeries row is a dict. Blank means auto: value, then out0, then close.

Putting close in source does not compute the indicator on closes. It sends close to the history fetch, which fails with a history error that names neither field.

Reading the result

{{rsi.latest.value}}      value on the last bar the feed returned
{{rsi.previous.value}}    value one bar earlier
{{rsi.at_offset.value}}   value offsetBars back
{{rsi.series}}            array of the last tailBars values
{{rsi.series[0].value}}   oldest value in that array
{{rsi.outputs}}           list of output names
{{rsi.bars_used}}         how many bars were actually computed over

Which bar is latest

latest is the last bar in the fetched history, and that is not always a closed one. Unlike barOffset and priorPeriodOhlc, the indicator node does not drop a still-forming bar:

• If the feed has already opened the current bar, latest is that forming bar and it repaints for the rest of the period. previous is then the bar that just closed.
• If the feed has not opened it yet, latest is the bar that just closed and previous is the one before it.

Which of the two you get is decided by when in the period the run happens. An interval schedule fires on the clock boundary plus FLOW_INTERVAL_ALIGN_OFFSET (default 2 seconds). Measured on a live instance at :02 past the minute, the feed had not yet opened the new minute's bar, so {{ind.latest.*}} was the candle that had just closed. That is what makes a schedule-driven, non-repainting strategy possible, and it is what Tutorial 12 relies on.

If a repainting latest is not acceptable and you would rather not depend on the timing, read previous instead and accept a one-bar lag.

FLOW_HISTORY_CACHE_TTL (default 30 seconds) is how long the indicator and history nodes reuse a fetch. That is half a 1-minute candle, so lower it to 2 or 3 for a 1-minute strategy or two runs a minute apart can see the same bars.

Single-output and tuple indicators

Single-output indicators (rsi, sma, ema, atr, …) expose value.

Tuple indicators, the ones whose library function returns several series, expose out0, out1, out2, … in the library's own order. They have no value key at all, so {{st.latest.value}} resolves to nothing: a varCondition reading it takes neither branch, an order field reading it fails the node, and a log message prints the braces back at you.

Indicator	out0	out1	out2
macd	MACD line	signal	histogram
bbands	upper	middle	lower
supertrend	level	direction (see below)	none
stochastic	%K	%D	none
adx	+DI	-DI	ADX
donchian	upper	middle	lower

So a supertrend level is {{st.latest.out0}} and its direction is {{st.latest.out1}}. {{st.outputs}} lists the names a given indicator actually produced, which is the quickest way to check one you have not used before.

Supertrend's direction is inverted

out1 does not follow the usual convention:

out1	Meaning
-1	Uptrend, price is above the supertrend line
+1	Downtrend, price is below the supertrend line

So "is it in an uptrend" is &#123;&#123;st.latest.out1&#125;&#125; < 0, not > 0. A strategy written on the usual assumption trades backwards: it buys every downtrend and sells every uptrend, and nothing in the run log looks wrong.

Verified two ways: against 497 real 1-minute bars, and by checking the library directly, where every bar with out1 == -1 has its close above out0 and every bar with out1 == +1 has its close below it.

Historical indicator values

Use offsetBars to read a specific bar:

{ "data": { "indicatorName": "supertrend",
            "params": "{\"period\": 10, \"multiplier\": 3}",
            "offsetBars": 5, "outputVariable": "st" } }

{{st.at_offset.out0}} is the supertrend level five bars back from latest. offsetBars: 0 puts at_offset on the same bar as latest, and offsetBars: 1 on the same bar as previous, so the three readers overlap by design: latest and previous are the fixed pair, at_offset is the one you move.

You can reverse-index series instead, but prefer at_offset: series[0]'s meaning depends on tailBars, so raising tailBars from 6 to 10 silently changes which bar series[0] refers to. at_offset is unambiguous.

Nesting indicators

Set sourceSeries to another indicator's series to compute an indicator of an indicator, for example a 9-period SMA of RSI:

{ "id": "smoothed", "type": "indicator", "position": { "x": 300, "y": 100 },
  "data": { "indicatorName": "sma", "params": "{\"period\": 9}",
            "sourceSeries": "{{rsi.series}}", "tailBars": 5,
            "outputVariable": "rsiSma" } }

Give the upstream node a large enough tailBars to feed the nested window. A 9-period SMA needs at least 9 upstream values.

Only single-series indicators can be nested (sma, ema, rsi, wma, stdev, highest, lowest, …). Anything needing independent high/low/close inputs (atr, supertrend, adx, …) cannot be reconstructed from one collapsed output series and returns a clear error.

sourceSeries also accepts a raw history array, {{h.data}} uses each row's close. Override with sourceField to use high, low, or a specific out1.

Crossovers

crossover, crossunder, and cross are not available as an indicator node: they need two independent series, and this node reads one symbol. Build a crossover from two indicator nodes plus an andGate, see Tutorial 3.

correlation and beta are excluded for the same reason (they compare two symbols).

The complete exclusion list

tradeboard.ta ships 127 public functions, more than the indicator node exposes. Eight need a second series and three cannot be computed by the installed build, leaving the 116 you can use:

Excluded	Why	What to do instead
crossover, crossunder, cross	Compare two series	Two indicator nodes plus an andGate, Tutorial 3
correlation, beta	Compare two symbols	Not available in Flow; use the Python Strategy Host
exrem, flip, valuewhen	Need a second boolean series, and carry state across bars	Restructure as a stateless condition, or use the Python Strategy Host
median_bands, ulcerindex, vi	The installed tradeboard build returns no usable values for these	Not available; selecting one fails the run with an explicit message

Everything else in the library works as an indicator node.

Available indicators

Category	Functions
Trend	sma, ema, wma, dema, tema, hma, vwma, alma, kama, zlema, t3, frama, trima, mcginley, vidya, alligator, ma_envelopes, supertrend, ichimoku, ckstop
Momentum	rsi, macd, stochastic, stochf, stochrsi, cci, williams_r, bop, elderray, fisher, crsi, cmo, trix, mom, apo, ppo, po, dpo
Volatility	atr, natr, true_range, bbands, bbpercent, bbwidth, keltner, donchian, chaikin, rvi, ultimate_oscillator, uo_oscillator, massindex, chandelier_exit, hv, starc
Volume	obv, obv_smoothed, vwap, mfi, adl, cmf, emv, force_index, nvi, nvi_with_ema, pvi, pvi_with_signal, volosc, vroc, kvo, pvt, rvol
Oscillators	roc, rocp, rocr, rocr100, awesome_oscillator, accelerator_oscillator, aroon_oscillator, cho, chop, kst, tsi, stc, gator_oscillator, coppock
Statistical	linreg, linregangle, linregintercept, lrslope, variance, tsf, median, mode
Hybrid	adx, adxr, dx, dmi, minus_dm, plus_dm, aroon, pivot_points, psar, fractals, rwi
Price transform	avgprice, medprice, midprice, midpoint, typprice, wclprice
Utility	highest, lowest, change, stdev, rising, falling

Parameter names follow the tradeboard.ta signatures, {"period": 14} for most, {"fast_period": 12, "slow_period": 26, "signal_period": 9} for MACD, {"period": 10, "multiplier": 3} for Supertrend. Required parameters get sensible defaults if you omit them.