Skip to main content

Prompt Library

Manage reusable prompts for AI nodes across workflows, ensuring consistent AI behavior.

What is the Prompt Library?​

A central repository for:

  • System prompts
  • Instruction templates
  • Few-shot examples
  • Tested prompt patterns

Benefits​

  • Consistency: Same prompt behavior everywhere, across workflows
  • Versioning: Track prompt changes with version control
  • Collaboration: Share prompts across the team
  • Testing: Easier A/B testing
  • Easy updates: Change once, applied everywhere

Creating a Prompt​

  1. Navigate to Automation → Prompt Library
  2. Click + Create Prompt (or + New Prompt)
  3. Enter:
    • Name - Descriptive identifier
    • Description - What it does
    • Category - Organization
    • Content - The prompt text
  4. Save

Prompt Structure​

System Prompt​

Sets AI behavior and personality:

You are a helpful customer support agent for Expedify.
You are friendly, professional, and concise.
Always verify information before providing answers.
If unsure, ask clarifying questions.

Instruction Template​

Specific task instructions with variables:

Analyze the following customer message and:
1. Identify the main issue
2. Determine sentiment (positive/neutral/negative)
3. Suggest a response

Customer message: {{message}}

Respond in JSON format:
{
  "issue": "...",
  "sentiment": "...",
  "suggested_response": "..."
}

With Variables​

Use placeholders for dynamic content:

Summarize the following customer information:

Name: {{contact_name}}
Company: {{company_name}}
Recent Activity: {{activity_summary}}

Provide a brief 2-3 sentence summary focusing on their needs.

Using Variables​

Insert dynamic content:

VariableSource
{{contact.name}}CRM contact
{{input.message}}Workflow input
{{kb_results}}KB search results
{{previous_output}}Prior node output

Override variables to pass different values each execution:

  • From trigger data
  • From previous nodes
  • From CRM fields

Using in Workflows​

LLM Node​

  1. Add LLM node to workflow
  2. In the prompt field, click Library (or select Use Prompt from Library)
  3. Select your prompt
  4. Map variables to inputs

Combining Prompts​

Use multiple prompts together:

System: {{prompt:customer_support_agent}}

Context: {{prompt:product_knowledge}}

Task: {{prompt:response_generator}}

User message: {{input.message}}

Prompt Categories​

Organize by purpose:

CategoryExamples
SystemAgent personalities
AnalysisSentiment, classification
GenerationEmails, summaries
ExtractionData parsing
ConversationChat responses

Version Control​

  • Each save creates a version
  • View version history
  • Revert to previous versions
  • Compare versions side-by-side

Creating Versions​

  1. Open prompt
  2. Make changes
  3. Click Save as New Version
  4. Add version notes

Comparing Versions​

  1. Click History
  2. Select versions to compare
  3. View diff

Rolling Back​

  1. Find previous version
  2. Click Restore
  3. Confirm

Testing Prompts​

Quick Test​

  1. Open prompt
  2. Click Test
  3. Enter sample input
  4. View output
  5. Iterate

A/B Testing​

Compare prompt versions:

  1. Create variant prompts
  2. Use in AI Evaluation
  3. Compare metrics
  4. Select winner

Best Practices​

  1. Be specific - Clear instructions
  2. Use examples - Show expected output
  3. Version everything - Track changes
  4. Test thoroughly - Multiple scenarios
  5. Document purpose - Future reference

Example Prompts​

Email Summary​

Summarize this email thread in 2-3 bullet points:
- Key points discussed
- Action items
- Next steps

Email thread:
{{email_thread}}

Lead Qualification​

Analyze this lead and score from 1-100:

Lead info:
- Company: {{company.name}}
- Size: {{company.size}}
- Industry: {{company.industry}}
- Behavior: {{visitor.page_views}} page views

Scoring criteria:
- Company fit (40 points)
- Engagement level (30 points)
- Budget signals (30 points)

Output JSON: {"score": X, "reasoning": "..."}

Related & Next Steps​

Last updated on