For the complete documentation index, see llms.txt. This page is also available as Markdown.

Flow Editor

Flow is OpenAlgo's node-graph strategy builder at /flow. You wire market data into indicators, indicators into conditions, and conditions into order execution, no Python required.

This section is the technical reference. It assumes you are building a real strategy and want to know exactly what each node does, what the execution model guarantees, and where the boundaries are.


What Flow can do

Capability
Nodes

Trigger on a schedule, a webhook, a price level, or an order fill

start, webhookTrigger, priceAlert, orderUpdateTrigger

Run a schedule on the clock, inside a configurable market-hours window

start (marketHoursOnly, marketHoursStart/End/Exchange)

Read live quotes and market depth

getQuote, multiQuotes, getDepth, subscribeLtp/Quote/Depth, unsubscribe

Read historical OHLCV at any supported timeframe

history, barOffset, priorPeriodOhlc

Compute any of 116 technical indicators

indicator

Branch on price, time, position, funds, or any computed value

priceCondition, timeWindow, timeCondition, positionCheck, fundCheck, varCondition

Combine conditions

andGate, orGate, notGate

Place equity, futures, and options orders

placeOrder, smartOrder, optionsOrder, optionsMultiOrder, basketOrder, splitOrder

Drive every order field from the alert that triggered the run, not just symbol

any order node, using {{webhook.*}} on exchange, action, quantity, product, priceType, price, triggerPrice, offset, optionType, expiryType, positionSize, splitSize

Manage open orders and positions

modifyOrder, cancelOrder, cancelAllOrders, closePositions

Resolve option symbols, expiries, and chains, by relative expiry or an exact DDMMMYY date

optionSymbol, expiry, optionChain, syntheticFuture

Look up contract details and supported intervals

symbol, intervals

Alert and log

telegramAlert, whatsappAlert, log, httpRequest

Track P&L for one strategy, not the account

strategyPnl

Read the account: orders, trades, positions, holdings, funds, margin

orderBook, tradeBook, positionBook, openPosition, holdings, funds, margin, getOrderStatus

Check exchange holidays and market timings

holidays, timings

Detect a new day, week, month, quarter or year

calendar

Pause within a run

delay, waitUntil

Arithmetic and state within a run

mathExpression, variable

61 node types in total, every one documented in the Node Reference, with the JSON each one expects in the JSON Format reference. Every order node calls the same service functions as /api/v1/, so Analyzer (sandbox) mode, Action Center approval, and Telegram or WhatsApp alerts all behave identically to an API-placed order.

A workflow can be checked before it is imported. The flow-builder skill in the OpenAlgo repo runs the importer's own validator and then flags the mistakes that import cleanly and misbehave later, chiefly a data key that no node reads:

What Flow cannot do

These are hard boundaries, not oversights. Limitations covers each in detail with workarounds.

Limitation
Practical effect

No state across runs

variable values reset every run. "Have I already entered today?" must be answered from the broker (positionCheck, orderBook), not a counter. Period boundaries are the exception: calendar answers "is a new week/month/quarter" from the exchange calendar, no memory needed.

delay and waitUntil block the run

They hold the execution slot, and for a webhook the triggering HTTP request too, for their full duration. delay is silently capped at 300 seconds and then continues, so a workflow written as "wait 30 minutes" exits after five. waitUntil is capped at 30 minutes and errors beyond that rather than waiting. Waiting out a session means a second workflow on its own schedule.

No loops

You cannot iterate a symbol list. One workflow handles one symbol; clone it per symbol, or drive one workflow from {{webhook.symbol}}, see Tutorial 11.

No backtesting

Flow runs forward only. Use the Python Strategy Host or the openalgo SDK to backtest.

No pandas objects

Variables are JSON. You get arrays of records and single values, not a pandas.Series.

Max 200 bars per fetch

Deep history (10 years of 1-minute data) is refused by design. See Market Data.

Triggers cannot read upstream data

A trigger is the entry point, so its fields cannot contain {{variables}}.

Indicators need one symbol

correlation, beta, crossover, crossunder need two independent series and are not available as a single indicator node. Crossovers are built from two indicator nodes, see Tutorial 3.

closePositions ignores its filter in live trading

It honours symbol, exchange and product in Analyzer mode and squares off the entire book live, so a filtered node looks correct in sandbox testing. Close a specific leg with smartOrder and positionSize: 0 instead.

An unknown data key is silently ignored

strikeOffset where the field is offset imports cleanly, runs successfully, and leaves the node on its default. Nothing in the run says so. Check the workflow before importing it.

Live vs Analyzer mode

Flow does not have its own paper-trading switch. It inherits the global Analyzer toggle:

  • Analyzer on: orders are simulated by the Sandbox engine. Order ids are real-looking, mode is analyze, and nothing reaches your broker.

  • Analyzer off: orders go to the live broker.

Always build and validate a strategy with Analyzer on. Every run log quoted in this section was captured in Analyzer mode against a live broker connection.

One thing behaves differently between the two, and it is the one place where Analyzer testing can mislead: closePositions honours its symbol, exchange and product filter in Analyzer mode and ignores it live.

Last updated