Canvas Workflow Overview

Learn the core concepts, node system, and data flow of BeeQuant canvas workflows.

What is a canvas workflow?

A BeeQuant canvas workflow is a visual quantitative strategy-building system. By dragging nodes onto the canvas and connecting data flows, you can build a complete quantitative trading strategy without writing code.

A workflow consists of seven core node types. Each node performs a specific function, and connections pass data between nodes to form a complete signal-processing pipeline from data acquisition to trade execution.

No-code building

Configure strategies visually with drag and drop—no programming background required.

Modular design

Configure each node independently and combine them flexibly for different strategy requirements.

AI-native support

Use built-in AI models including LightGBM, LSTM, Transformer, and LLMs.

Data flow

Data flows through the workflow in the following order. Each node processes its input and passes the result downstream:

Data source
Data processing
[AI model]
Strategy type
Position management
Risk control
Trading engine
Tip

The AI model node is optional, as indicated by the square brackets above. Traditional technical-indicator strategies that do not require machine-learning predictions can connect the data processing node directly to the strategy type node.

Node connection rules

Connections must follow specific rules so data can flow correctly. The table below lists the connection constraints for each node:

NodeUpstream node (input)Downstream node (output)Description
Data sourceNone (starting point)Data processingEvery workflow must begin with a data source node.
Data processingData sourceAI model / Strategy typeConnect directly to a strategy node or route the data through an AI model first.
AI modelData processingStrategy typeOptional node for machine-learning predictions.
Strategy typeData processing / AI modelPosition managementGenerates trading signals from the data.
Position managementStrategy typeRisk controlControls capital allocation and take-profit/stop-loss rules.
Risk controlPosition managementTrading engineMonitors overall risk and triggers forced liquidation.
Trading engineRisk controlNone (end point)Every workflow must end with a trading engine node.
Port color-matching rules

Connection ports use different colors. Only ports of the same color can be connected; the system automatically validates every connection.

Cyan
Data source → Data processing, carrying raw OHLCV data
Teal
Data processing → AI model → Strategy type, carrying factor features
Pink
Risk control → Trading engine, carrying execution instructions
Required nodes

A complete strategy workflow requires all of the following:

  • Data source → data origin
  • Data processing → feature calculation
  • Strategy type → signal generation
  • Position management → capital control
  • Risk control → risk management
  • Trading engine → order execution
Optional nodes

Add the following node only when required by the strategy:

  • AI model → add when the strategy requires machine-learning predictions
Tip: Traditional technical-indicator strategies, such as moving-average crossovers and MACD, do not need an AI model node. Connect data processing directly to the strategy type node.

Node details

Data source node

A required parameter-definition card for every data branch

The data source node is a parameter-definition card. It does not call exchange APIs, retrieve data, or output a DataFrame. It only defines the extraction parameters required by the downstream data extraction and processing node. It supports the two major exchanges Binance and OKX, covering spot, USDT-margined futures, and coin-margined futures markets.

Time range
Defines the data window for backtesting or live trading
Exchange and market
Selects the data source and market type
Trading pair
Selects the instrument, such as BTCUSDT
Candle interval
1m, 5m, 15m, 1h, 4h, 1d, etc.

Data processing node

The feature-engineering hub for technical indicators and quantitative factors

The data processing node is the strategy's feature-engineering hub and the only node in the entire workflow that actually extracts OHLCV data. It reads the upstream data-source parameters, retrieves raw OHLCV data, and transforms it into factor features for strategy decisions or AI model training. Use the flexible expression syntax to define technical indicators.

Lookback data
Sets the amount of history required for factor calculations
Factor configuration
Defines technical indicator and quantitative factor expressions
Label definition
Defines target variables for AI model training
Conditional filtering
Removes rows that do not meet specified conditions

AI model nodes

Optional nodes that add machine-learning predictions

An AI model node uses the factors emitted by data processing as features to predict future returns or price direction. Its predictions can drive asset-selection rankings or timing signals.

Task types: Every AI model supports ranking for asset selection, binary classification for up/down direction, and regression for return prediction.

Strategy groupStrategy type + Position management + Risk control

Strategy type, position management, and risk control form a bound strategy group, shown on the canvas inside a dashed outline. They must be connected in order and together implement the complete strategy logic from signal generation through risk management.

① Strategy type
Generates entry and exit signals
② Position management
Controls capital allocation and take-profit/stop-loss rules
③ Risk control
Monitors account risk and triggers liquidation protection

Strategy type node

The core signal generator that defines when and what to trade

The strategy type node is the core trading-signal generator. It supports two main strategy classes: timing strategies, which use conditional expressions to decide when to enter and exit, and selection strategies, which rank factors to choose instruments.

Timing strategies
  • • Use conditional expressions to determine entries and exits
  • • Best suited to one or a small number of instruments
  • • Support long entry, short entry, long exit, and short exit signals
  • • Can be combined with grid mode for layered position additions
Selection strategies
  • • Rank factors or model scores to select instruments
  • • Periodically rerank and rebalance the selected instruments
  • • Three subtypes: selection long, selection short, and long/short hedge
  • • Support equal-weight and top-heavy allocation

Position management node

Controls capital allocation, entry size, and take-profit/stop-loss rules

Position management is central to strategy execution. It controls critical parameters such as capital allocation, entry size, and take-profit and stop-loss rules, directly shaping the strategy's risk-and-return profile.

Normal mode
  • • Fixed-percentage entries control risk per trade
  • • Supports fixed take-profit/stop-loss and trailing stops
  • • Available for timing and selection strategies
Grid mode
  • • Adds positions in layers to average the entry cost
  • • Supports ranging and trending modes
  • • Available only for timing strategies

Risk control node

The final safety barrier protecting your capital

The risk control node is a quantitative strategy's final safety barrier. It monitors runtime NAV—including realized and unrealized P&L while isolating pure deposits and withdrawals—using mutually exclusive initial-capital or high-water-mark drawdown modes, and forces all positions to close when the threshold is reached.

Drawdown liquidation
  • • Runtime NAV includes realized and unrealized P&L
  • • Initial-capital and high-water-mark modes are mutually exclusive
  • • Defaults: 20% from initial capital; 25% from the high-water mark
Cooldown
  • • Pauses strategy trading after liquidation
  • • Waits for market conditions to stabilize
  • • Helps avoid consecutive losses and emotional trading

Trading engine node

The execution terminal that converts strategy signals into orders

The trading engine is the strategy's execution terminal, responsible for turning strategy signals into actual orders. It supports backtest mode for historical validation and live mode for real trading.

Run mode
Backtest validation / Live trading
Exchange settings
Multiple exchanges and markets
Capital settings
Initial capital denominated in USDT
Trading parameters
Fees, slippage, and leverage

Workflow examples

The following two typical strategy workflows show how nodes are connected on the canvas:

Example 1: Traditional technical-indicator strategy (without an AI model)

Suitable for timing or selection strategies based on moving averages, MACD, RSI, and other technical indicators
Flow: Data source → Data extraction & processing → Strategy type → Position management → Risk control → Trading engine. The data processing node connects directly to the strategy type node, and factor expressions define the trading conditions.

Example 2: AI prediction strategy (with an AI model)

Suitable for strategies that require machine-learning predictions, such as instrument ranking and return forecasting
Flow: Data source → Data extraction & processing → LightGBM model → Strategy type → Position management → Risk control → Trading engine. The AI model node receives factor features and emits predictions for the strategy node.

Next steps

Now that you understand the basics of canvas workflows, continue in the following order:

    BeeQuant - AI Quantitative Trading Platform