Skip to main content

Prompting Tips

Get better results from TalkingSchema's AI copilot with these tips and example prompts.

Be specific about field types​

Instead of: "Add a carbon score field to products" Try: "Add a carbon_intensity_score field to products as DECIMAL(5,2), not null, default 0"

Reference your domain​

The more context you give about your industry and use case, the better the AI's suggestions:

"I'm building a global sustainable supply chain system. Create tables for suppliers with carbon tier ratings (A/B/C/D), products with carbon intensity scores, warehouses, purchase orders, and shipment emissions tracking."

Describe relationships explicitly​

Instead of: "Connect suppliers and purchase orders" Try: "Each purchase order belongs to one supplier and one destination warehouse. Purchase orders have many line items, each referencing a product with a quantity and unit price."

Use industry-standard terms​

The AI understands: supply chain, 3NF, audit log, soft delete, polymorphic relationship, surrogate key, composite key, junction table, lookup table, carbon offset, emissions tracking

Ask the AI to explain​

"Explain the relationship between the purchase_orders and purchase_order_items tables you just created"

"What indexes would you recommend for querying shipments by carrier and delivery date?"

"Is this schema normalized? Where might I have data redundancy?"

Iterate incrementally​

Build complex schemas step by step:

Step 1: "Create a core supply chain schema with suppliers, products, and warehouses"
Step 2: "Add purchase orders and sales orders with status enums"
Step 3: "Add shipments table with carrier and emissions_kg tracking"
Step 4: "Add carbon_offset_credits table linked to suppliers"

Common schema patterns​

Supply chain core​

"Create a supply chain schema with suppliers (carbon_tier A/B/C/D), products (carbon_intensity_score), warehouses, purchase orders, and shipments. Use snake_case. All tables need created_at."

Carbon tracking​

"Add a carbon_offset_credits table linked to suppliers. Include tonnes_co2, vintage_year, verified_by, and certification_url. Add a carbon_tier enum (A, B, C, D) to the suppliers table."

Inventory management​

"Add an inventory table that tracks quantity per product per warehouse. Include updated_at and a low_stock_threshold field."

Order fulfillment​

"Design a full order fulfillment flow: purchase_orders → purchase_order_items → products → shipments. Include an order_status enum: pending, confirmed, processing, shipped, delivered, cancelled."

What the AI can't do​

  • Connect to your actual database to check existing data
  • Run queries against your database
  • Generate application code (only SQL/DBML schema output)
  • Guarantee zero breaking changes when modifying existing schemas