HomeGuidesAnnouncementsCommunity
Guides

Prompt Composer

Suggest Edits

The Prompt Composer node is a specialized component that dynamically assembles prompts by combining base instructions, relevant examples, and contextual data. It serves as a bridge between your prompt library and runtime execution, enabling context-aware, dynamic prompt generation.

Key Features

  • Dynamic prompt assembly
  • Example matching based on user input
  • Template context injection
  • Configurable matching thresholds
  • Integration with prompt library

Configuration

Basic Setup

  1. Add the Prompt Composer node to your skill
  2. Select a prompt from your prompt library
  3. Configure template context if using dynamic data
  4. Set up match values for example selection

Template Context

Template context allows you to inject dynamic data into your prompts:

Base text: "Here is the sales data: {{dataset_query.df}}"

The variable in double curly braces will be replaced with actual data at runtime.

Match Values

  • Create variables to match against example triggers
  • Standard format: user__full_query
  • Enable "Include as chat parameter" for user input matching
  • Multiple match values can be configured for complex matching scenarios

Node Outputs

composed_prompt

  • Complete assembled prompt including:
    • Base prompt text
    • Matched examples
    • Injected template data
  • Ready for immediate use by Prompt Runner

matched_examples

  • List of examples that matched the query/match term
  • Useful for debugging and optimization

Best Practices

Threshold Configuration for Prompt

  • Default: 0.4 (recommended starting point)
  • Higher thresholds (0.8+): More precise matching
  • Lower thresholds (0.1-0.3): More flexible matching
  • Adjust based on your use case needs

Debugging

  1. Check template context injection (is df or other input being included in the composed prompt?)
  2. Verify example matching
  3. Review final composed prompt
  4. Adjust prompt match thresholds if needed

Common Issues and Solutions

No Examples Matching

  • Lower the matching threshold
  • Review match value configuration
  • Check example trigger phrases
  • Verify user input format

Use Cases

1. Brand Analysis Assistant

  • Purpose: Analyze brand performance and characteristics
  • Implementation:
    • Base prompt with brand analysis instructions
    • Examples for different analysis types (sales, perception, growth)
    • Template context set to a df with brand performance data
    • Match values based on the type of analysis that answers the user's request

2. Customer Support Router

  • Purpose: Route and format customer inquiries
  • Implementation:
    • Base prompt with response guidelines
    • Examples for different support categories
    • Template context with conversation history
    • Match values based on inquiry type or themes

3. Market Analysis Reporter

  • Purpose: Generate market analysis reports
  • Implementation:
    • Base prompt with analysis framework
    • Examples for different market conditions
    • Template context with market data
    • Match values based on analysis type

Integration Examples

With Dataset Query

Dataset Query → Prompt Composer → Prompt Runner

Useful for data-driven analysis and reporting

With User Input Processing

User Input → Input Processor → Prompt Composer → Prompt Runner

Ideal for interactive applications

With Multiple Data Sources

Dataset Query 1 ↘
Dataset Query 2 → Prompt Composer → Prompt Runner
Dataset Query 3 ↗

Perfect for comprehensive analysis scenarios

Updated 8 days ago