Flow Builder

7 min read

Overview

The Flow Builder is Cacuda's visual drag-and-drop pipeline editor. It lets you design complex moderation workflows by connecting nodes on a canvas — no code required.

The Canvas

The flow builder interface has three panels:

  • Node Palette (left) — drag nodes from here onto the canvas
  • Canvas (center) — the main workspace where you build your flow by connecting nodes
  • Config Panel (right) — click any node to configure its settings here

Node Types

Nodes are color-coded by category:

Triggers (Blue)

Entry points for your pipeline. Webhook, WordPress, Drupal, and MCP triggers receive incoming content and start the flow.

Processors (Amber)

Analyze content and produce a pass/fail result. Includes ML detectors (spam, NSFW, toxicity), word list checks, text length checks, and regex matchers.

Actions (Green / Red)

Terminal nodes that take action on the content. Approve (green) publishes content; Reject (red) removes or flags it. Can trigger callbacks to the originating platform.

Conditions (Indigo)

Routing nodes that branch the flow based on scores, field values, or comparison operators. Use these to build complex decision trees.

Connecting Nodes

To connect two nodes, click and drag from an output handle on the source node to an input handle on the target node.

Processor nodes have two output handles:

  • Pass (green) — content scored below the threshold
  • Fail (red) — content scored above the threshold

Conditional Branching

Use Condition nodes to route content based on dynamic values:

  • Score-based — route based on detection scores (e.g., if spam score > 0.8, escalate for manual review)
  • Field-based — check metadata fields (e.g., if source = "vip-user", skip detection)
  • Comparison operators — equals, not equals, greater than, less than, contains

Configuring Nodes

Click any node on the canvas to open its configuration in the right panel. Common settings include:

  • Thresholds — set the score cutoff for pass/fail on detector nodes
  • Callback URLs — where to send approve/reject results
  • Word lists — select which word list to check against
  • Content field mapping — specify which payload field contains the content to analyze

Testing

Click the Test button to dry-run your pipeline with sample data. The test view shows:

  • Step-by-step execution through each node
  • Input and output data at each stage
  • Detection scores and pass/fail decisions
  • The final action taken
Tip: Test mode uses real ML models but does not trigger callbacks or count toward your execution history.

Activating

When your pipeline is ready, click Save to persist changes, then Activate to enable live processing. Active automods process incoming webhooks in real-time.

You can Pause an active automod at any time without losing your configuration. Paused automods reject incoming webhooks with a 503 status.

Credit Costs

Only ML detection nodes consume credits. Rule-based checks are free:

Node TypeCredit Cost
Spam Detection1 credit
NSFW Text Detection1 credit
NSFW Image Detection2 credits
Toxicity Detection1 credit
Word List CheckFree
Regex MatchFree
Text Length CheckFree
ConditionFree
Tip: Place free checks (word lists, regex, length) before paid ML detection nodes to minimize credit usage.