Parent Epic

Part of #6991 - AI Agent Registry

Description

Add support for managing schemas related to AI/ML models, including model input/output definitions, prompt templates, and RAG pipeline configurations.

Background

As AI becomes integral to enterprise applications, there's a need to govern:

• Model input/output schemas (what data goes in, what comes out)
• Prompt templates and their versions
• RAG (Retrieval-Augmented Generation) pipeline configurations
• Feature store schemas for ML pipelines

Requirements

New Artifact Types

1. MODEL_SCHEMA

   • Input schema definition (JSON Schema or similar)
   • Output schema definition
   • Model metadata (version, provider, capabilities)

2. PROMPT_TEMPLATE

   • Template content with variable placeholders
   • Variable schema definition
   • Version history for prompt iteration

3. RAG_CONFIG (optional, future)

   • Embedding model configuration
   • Vector store settings
   • Retrieval parameters

Features

• Validation of model schemas
• Version comparison for prompt templates
• Compatibility checking between model versions
• Integration with feature store schemas

Acceptance Criteria

• MODEL_SCHEMA artifact type works end-to-end
• PROMPT_TEMPLATE supports versioning
• UI displays AI-related artifacts appropriately
• Documentation covers AI schema use cases

Priority

P1 - High priority as part of AI differentiation strategy

References

• Model Cards for Model Reporting
• Prompt Engineering Best Practices