Prompt Library Configuration

Overview

LLMling's prompt library allows defining reusable prompts that can be shared across agents. Prompts are defined in the prompts section of your configuration and can be referenced by name.

Basic Structure

prompts:
  # Main system prompts for defining agent behavior
  system_prompts:
    expert_analyst:
      content: |
        You are an expert data analyst.
        Focus on finding patterns and insights.
      category: role

    step_by_step:
      content: |
        Break tasks into sequential steps.
        Explain each step thoroughly.
      category: methodology

# Using prompts in agents
agents:
  analyst:
    system_prompts:
      # Direct string prompts
      - "You help with analysis."

      # Reference library prompts
      - type: library
        reference: expert_analyst
      - type: library
        reference: step_by_step

Prompt Categories

System Prompts

Define core agent behaviors and methodologies:

prompts:
  system_prompts:
    # Role definition
    technical_writer:
      content: |
        You are an expert technical writer.
        Focus on clarity and precision.
        Use proper terminology consistently.
      category: role

    # Methodology definition
    step_by_step:
      content: |
        Follow these steps for each task:
        1. Understand requirements
        2. Plan approach
        3. Execute systematically
        4. Verify results
      category: methodology

    # Tone/style definition
    professional:
      content: |
        Maintain formal, business-appropriate language.
        Be concise but thorough.
      category: tone

Prompt Types

The type field categorizes prompts by their purpose:

Role Types

Define WHO the agent is:

system_prompts:
  expert_dev:
    category: role
    content: |
      You are a senior software developer with expertise in:
      - System architecture design
      - Code quality assessment
      - Performance optimization

  data_scientist:
    category: role
    content: |
      You specialize in data analysis and machine learning.
      Your expertise includes statistical modeling and data visualization.

Methodology Types

Define HOW the agent works:

system_prompts:
  analytical:
    category: methodology
    content: |
      Approach problems systematically:
      1. Gather and analyze data
      2. Identify patterns and trends
      3. Form evidence-based conclusions
      4. Present findings clearly

  iterative:
    category: methodology
    content: |
      Work in small iterations:
      - Start with minimal viable approach
      - Test and validate results
      - Refine based on feedback
      - Scale successful patterns

Tone Types

Define communication STYLE:

system_prompts:
  formal:
    category: tone
    content: |
      Use professional, business-appropriate language:
      - Maintain formal tone
      - Be precise and clear
      - Avoid colloquialisms

  friendly:
    category: tone
    content: |
      Be approachable and helpful:
      - Use warm, welcoming language
      - Show empathy and understanding
      - Encourage questions and dialogue

Format Types

Define output STRUCTURE:

system_prompts:
  markdown:
    category: format
    content: |
      Format responses using Markdown:
      - Use headers for sections (# ## ###)
      - Use bullet points for lists
      - Use code blocks for code examples
      - Use tables for structured data

  structured:
    category: format
    content: |
      Structure responses with:
      1. **Summary** - Brief overview
      2. **Details** - Comprehensive explanation
      3. **Examples** - Practical illustrations
      4. **Next Steps** - Recommended actions

Using Library Prompts

Reference prompts in agent configuration:

agents:
  technical_assistant:
    model: gpt-4
    system_prompts:
      # Direct string prompts
      - "You are a technical assistant."
      - "Focus on helping with code and documentation."

      # Library reference prompts
      - type: library
        reference: technical_writer    # Role
      - type: library
        reference: step_by_step       # Methodology
      - type: library
        reference: professional       # Tone
      - type: library
        reference: markdown          # Format

  data_analyst:
    system_prompts:
      - type: library
        reference: expert_analyst
      - type: library
        reference: analytical
      - type: library
        reference: formal

Advanced Prompt Management

Dynamic Prompt Generation

Use functions to generate context-aware prompts:

prompts:
  system_prompts:
    context_aware:
      category: role
      content: |
        You adapt to user context and preferences.

agents:
  adaptive_agent:
    system_prompts:
      - type: library
        reference: context_aware
      - type: function
        function: "prompts:generate_user_context"
        arguments:
          user_type: "developer"
          experience_level: "intermediate"

Complete Example

prompts:
  system_prompts:
    # Roles
    technical_expert:
      category: role
      content: |
        You are a technical expert specializing in:
        - Software development best practices
        - System architecture and design
        - Code review and quality assurance

    code_reviewer:
      category: role
      content: |
        You are an experienced code reviewer focused on:
        - Code quality and maintainability
        - Security best practices
        - Performance optimization

    # Methodologies
    systematic:
      category: methodology
      content: |
        Follow this systematic approach:
        1. Understand requirements fully
        2. Break down complex problems
        3. Apply best practices consistently
        4. Validate results thoroughly

    # Tones
    professional:
      category: tone
      content: |
        Maintain professional communication:
        - Use formal, precise language
        - Be respectful and constructive
        - Provide clear explanations

    # Formats
    structured:
      category: format
      content: |
        Structure responses with:
        1. **Overview** - Brief summary
        2. **Analysis** - Detailed examination
        3. **Recommendations** - Actionable advice
        4. **Examples** - Practical illustrations

agents:
  senior_dev:
    model: gpt-4
    description: "Senior developer specialized in code review and optimization"
    system_prompts:
      - "Specialize in Python and TypeScript development."
      - type: library
        reference: technical_expert
      - type: library
        reference: code_reviewer
      - type: library
        reference: systematic
      - type: library
        reference: professional
      - type: library
        reference: structured

  mentor:
    model: gpt-4
    description: "Programming teacher and mentor"
    system_prompts:
      - type: library
        reference: technical_expert
      - type: file
        path: "prompts/teaching_style.j2"
        variables:
          approach: "socratic"
          patience_level: "high"
      - type: function
        function: "education:generate_learning_context"
        arguments:
          student_level: "beginner"
          topic: "programming_fundamentals"

Organization Best Practices

File Structure

Keep prompts organized in separate files:

# prompts/roles.yml
prompts:
  system_prompts:
    technical_expert:
      category: role
      content: ...

# prompts/styles.yml
prompts:
  system_prompts:
    professional:
      category: tone
      content: ...

# agents.yml
INHERIT:
  - prompts/roles.yml
  - prompts/styles.yml

agents:
  my_agent:
    system_prompts:
      - type: library
        reference: technical_expert
      - type: library
        reference: professional

Naming Conventions

Use clear, descriptive names:

• Roles: expert_analyst, code_reviewer, technical_writer
• Methodologies: step_by_step, analytical, iterative
• Tones: professional, friendly, formal, casual
• Formats: markdown, structured, bullet_points

Documentation

Document your prompts:

prompts:
  system_prompts:
    expert_analyst:
      category: role
      content: |
        You are an expert data analyst with 10+ years experience.

        Core competencies:
        - Statistical analysis and modeling
        - Data visualization and reporting
        - Business intelligence and insights
      # Internal documentation (not sent to agent)
      description: "Primary role for data analysis agents"
      tags: ["data", "analysis", "expert"]
      version: "1.2"

System Prompts Format

The system prompts format allows you to mix different prompt types:

agents:
  my_agent:
    system_prompts:
      - "Direct prompt"
      - type: library
        reference: expert_role
      - type: library
        reference: professional_tone

This format provides: - Type safety: Clear discrimination between prompt types - Flexibility: Mix different prompt sources in one list - Extensibility: Support for multiple prompt types - Consistency: Unified approach across all prompt types