Prompt & GenAI Types

Prompt types support managed prompt engineering workflows for generative AI. Unlike other product type extensions that attach metadata at the product level, prompt metadata extends individual fields within a product — each field represents a distinct prompt.

Available Types

Prompt Collection

A managed collection of AI prompts, supporting both single-turn and multi-turn (chat) formats.

• Icon: MessageSquare · Color: #f97316 (orange)
• Created by: Manual registration

How Prompt Collections Work

A Prompt Collection is a data product where each field (row in the Schema tab) represents an individual prompt. The Schema tab is enhanced with:

• An Add Prompt button for creating new prompts
• A prompt editor modal for configuring each prompt's metadata
• Visual indicators showing model compatibility and output format

Prompt Field Metadata

Each prompt (field) in a collection carries the following metadata:

Message Sequence

The core of any prompt — the sequence of messages sent to the model:

[
  {"role": "system", "content": "You are a helpful data analyst..."},
  {"role": "user", "content": "Summarize the following dataset: {{data}}"},
  {"role": "assistant", "content": "Here is a summary..."}
]

Multi-turn prompts can include system, user, and assistant messages to demonstrate the expected conversation flow.

Template Variables

Define placeholder variables that are injected at runtime:

[
  {"name": "data", "type": "string", "description": "The dataset to analyze"},
  {"name": "format", "type": "enum", "options": ["json", "markdown", "csv"]}
]

Model Configuration

Property	Description	Example
Model Compatibility	Which LLM models this prompt works with	["gpt-4", "claude-3-sonnet", "gemini-pro"]
Output Format	Expected response format	json, markdown, text, csv
Temperature	Sampling temperature	0.7
Max Tokens	Maximum output token count	2048

Categorization

Property	Description	Example
Use Case Category	What kind of task this prompt performs	summarization, extraction, classification, generation
Includes	References to other prompts (composition)	["system_prompt_v2", "safety_guardrails"]

Examples

Provide example inputs and expected outputs for documentation and testing:

// Example inputs
[{"topic": "AI safety", "format": "markdown"}]

// Example outputs
[{"summary": "## AI Safety Overview\n\nAI safety encompasses..."}]

Tab Availability

Feature	Available?	Notes
Overview	✅	Standard overview
Schema	✅	Enhanced with prompt editor
Data Profiling	❌	Not applicable
Quality Health	✅	Standard quality checks
Lineage	✅	Track prompt dependencies
Governance	✅	Full governance suite
Versions	✅	Track prompt version history

Use Cases

Prompt Collections are ideal for:

• Prompt Libraries — Centralized, versioned libraries of tested prompts
• AI Agent Playbooks — Multi-turn conversation templates for chatbots and agents
• Evaluation Suites — Prompt collections with example I/O for systematic model evaluation
• Governance & Compliance — Auditable records of prompts used in production AI systems
• Team Collaboration — Shared prompt repositories with ownership, review, and change tracking