Nodes are the building blocks of AI workflows. Each node performs a specific task—reading data, processing it, calling AI models, or saving results. Connect nodes to build powerful automation pipelines.

Node Categories

🔵 Data Source Nodes

Start your workflow by reading data from various sources:

• Dataset Source - Read samples from a dataset with schema-based outputs
• API Input - Accept input from API requests when flow is deployed

🟢 Processing Nodes

Transform and process data as it flows through your pipeline:

• Website Mapper - Find and map website URLs from company names or domains
• Page Scraper - Extract content from web pages using CSS selectors
• HTML Text Extractor - Clean HTML and extract plain text
• Iterator (Array Splitter) - Split arrays into individual items for processing
• Array Flatten - Flatten nested arrays into a single array

🟡 AI Nodes

Process data with language models:

• Prompt Node - Run AI prompts (use prompts from the Prompt Engineering Platform)

🟣 Output Nodes

Save or return results:

• Dataset Sink - Write results to a dataset for storage and analysis
• API Output - Define response format when flow is deployed as API

Node Anatomy

Common Features

All nodes share these features:

• Input handles (left side) - Connect incoming data
• Output handles (right side) - Send data to next nodes
• Configuration panel - Click node to configure settings
• Status indicators - Blue (processing), green (completed), red (failed)
• Timing badges - Show execution time after running
• Enable/disable toggle - Power button to skip node without deleting
• Delete button - Remove node and its connections

Input/Output Pattern

Most processing nodes follow this pattern:

• Input: in - Generic input field (accepts any data structure)
• Output: out - Processed result
• Output: _input - Pass-through of original input

Dynamic Inputs/Outputs

Some nodes provide dynamic inputs or outputs based on configuration:

Dataset Source Node

Outputs include:

• out - Full dataset sample object
• _input - Pass-through
• out.field_name - Direct access to each schema field

Prompt Node

Inputs include:

• in - Generic input
• One input handle per prompt variable (e.g., url, content, context)

API Input Node

Outputs match your configured fields:

• out - Full input object
• out.field_name - Each configured field

Node Reference Pages

Detailed documentation for each node type:

Coming Soon

Learn more about upcoming nodes →

Best Practices

Node naming

• Nodes are auto-named based on type (e.g., prompt_1234567890)
• Use descriptive node types to make flows self-documenting
• Consider adding comments in the flow title/description

Error handling

• Failed nodes show red border and error icon
• Downstream nodes may still execute using cached data
• Use node disable feature to skip problematic nodes during debugging

Performance

• Check node timing badges to identify bottlenecks
• Prompt nodes with parallel iterations can process faster
• Web scraping nodes are typically the slowest (1-3 seconds per page)