---

name: schema-patterns description: Production-ready database schema patterns for AI applications including chat/conversation schemas, RAG document storage with pgvector, multi-tenant organization models, user management, and AI usage tracking. Use when building AI applications, creating database schemas, setting up chat systems, implementing RAG, designing multi-tenant databases, or when user mentions supabase schemas, chat database, RAG storage, pgvector, embeddings, conversation history, or AI application database. allowed-tools: Read, Write, Bash, Edit

Database Schema Patterns for AI Applications

Production-ready PostgreSQL/Supabase database schemas optimized for AI applications including chat systems, RAG (Retrieval-Augmented Generation), multi-tenancy, and usage tracking.

Instructions

1. Identify Required Pattern Type

Ask the user which schema pattern they need:

• chat: Conversation and messaging systems
• rag: Document storage with vector embeddings (pgvector)
• multi-tenant: Organization-based multi-tenancy
• user-management: Extended user profiles and metadata
• ai-usage: Token tracking, costs, and rate limiting
• complete: All patterns combined

2. Generate Schema

Use the generation script:

cd /home/vanman2025/Projects/ai-dev-marketplace/plugins/supabase/skills/schema-patterns
./scripts/generate-schema.sh <pattern-type> <output-file>

Pattern types: chat, rag, multi-tenant, user-management, ai-usage, complete

3. Validate Schema

Before applying, validate the generated schema:

./scripts/validate-schema.sh <schema-file>

This checks for:

• Proper table naming conventions (lowercase, underscores)
• Primary keys on all tables
• Foreign key relationships
• Index optimization
• pgvector extension usage (for RAG patterns)
• RLS policy structure
• Migration version format

4. Apply Migration

Apply the schema to your Supabase project:

./scripts/apply-migration.sh <schema-file> <migration-name>

This creates a timestamped migration file and validates before applying.

5. Seed Test Data (Optional)

For development, generate realistic test data:

./scripts/seed-data.sh <pattern-type>

Available Templates

Core Schemas

• chat-schema.sql: Complete chat/conversation system with users, conversations, messages, participants
• rag-schema.sql: RAG document storage with chunks, embeddings (pgvector), and similarity search
• multi-tenant-schema.sql: Organization-based multi-tenancy with orgs, teams, members, roles
• user-management-schema.sql: Extended user profiles, metadata, preferences
• ai-usage-tracking-schema.sql: Token usage, API costs, rate limiting, usage analytics

Supporting Templates

• migration-template.sql: Boilerplate migration structure with version tracking
• indexes-template.sql: Performance optimization index patterns
• rls-policies-template.sql: Row Level Security policy patterns

Key Features

pgvector Integration (RAG Schemas)

All RAG schemas include:

• Vector column setup with proper dimensions
• HNSW indexing for similarity search
• Cosine distance operators
• Automatic embedding column generation
• Metadata storage alongside embeddings

Multi-Tenancy Support

Organization-based isolation:

• Tenant identification (org_id on all tables)
• Team-based access control
• Member role management
• RLS policies for data isolation

Chat System Optimization

Optimized for real-time messaging:

• Conversation participants tracking
• Message ordering and pagination indexes
• Read/unread status tracking
• Typing indicators support
• Message search with full-text indexes

Performance Patterns

• Composite indexes for common queries
• Partial indexes for filtered queries
• Generated columns for computed fields
• Proper foreign key cascades
• Optimized join patterns

Examples

See the examples directory for:

• complete-ai-app-schema.md: Full schema combining all patterns
• migration-guide.md: Schema evolution and versioning
• indexing-strategy.md: Performance optimization guide

Best Practices

1. Always use lowercase with underscores for table/column names
2. Enable pgvector extension before creating vector columns
3. Add indexes on foreign keys for join performance
4. Use generated columns for computed fields (created_at, updated_at)
5. Implement RLS policies for security and multi-tenancy
6. Version all migrations with timestamps
7. Use halfvec for embeddings to save storage (16-bit vs 32-bit)
8. Add metadata JSONB columns for flexibility
9. Plan for soft deletes (deleted_at timestamp)
10. Include audit trails (created_by, updated_by)

Common Workflows

Setting Up a Chat Application

./scripts/generate-schema.sh chat schema.sql
./scripts/validate-schema.sh schema.sql
./scripts/apply-migration.sh schema.sql "initial-chat-schema"
./scripts/seed-data.sh chat

Building a RAG System

./scripts/generate-schema.sh rag schema.sql
./scripts/validate-schema.sh schema.sql
./scripts/apply-migration.sh schema.sql "add-rag-storage"
./scripts/seed-data.sh rag

Complete AI Platform

./scripts/generate-schema.sh complete schema.sql
./scripts/validate-schema.sh schema.sql
./scripts/apply-migration.sh schema.sql "complete-ai-platform"

Troubleshooting

pgvector not found: Enable the vector extension in Supabase dashboard (Database > Extensions)

RLS blocks queries: Check RLS policies or temporarily disable for testing (not recommended for production)

Slow similarity search: Ensure HNSW index is created on vector columns with proper operator class

Migration conflicts: Check migration version ordering and resolve conflicts manually

---

Skill Location: /home/vanman2025/Projects/ai-dev-marketplace/plugins/supabase/skills/schema-patterns/ Version: 1.0.0