Generate production database schemas from Excel and CSV with OpenAI and LangChain

Overview

This workflow automatically converts CSV or Excel files into a production-ready database schema using AI and rule-based validation.

It analyzes uploaded data, detects column types, relationships, and data quality, then generates a normalized schema. The output includes SQL DDL scripts, ERD diagrams, a data dictionary, and a load plan.

This eliminates manual schema design and accelerates database setup from raw data.

How It Works

1. File Upload (Webhook)

   • Accepts CSV or XLSX files via webhook endpoint
   • Initializes workflow configuration (thresholds, retry limits)

2. File Extraction

   • Detects file format (CSV or Excel)
   • Extracts rows into structured JSON
   • Merges extracted datasets

3. Data Cleaning & Profiling

   • Removes duplicates and normalizes values
   • Detects data types (integer, float, date, boolean, string)
   • Computes column statistics (nulls, uniqueness, distributions)
   • Generates file hash and sample dataset

4. Column Profiling Engine

   • Identifies potential primary keys
   • Detects cardinality and uniqueness levels
   • Suggests foreign key relationships based on value overlap

5. AI Schema Generation

   • Uses an AI agent to design normalized tables
   • Assigns SQL data types based on real data
   • Defines primary keys, foreign keys, constraints, and indexes

6. Validation Layer

   • Ensures schema matches actual data
   • Validates:
     • Data types
     • Primary key uniqueness
     • Foreign key overlap (>70%)
     • Constraint consistency
   • Detects circular dependencies

7. Revision Loop

   • If validation fails:
     • Sends feedback to AI agent
     • Regenerates schema
     • Retries up to configured limit

8. Schema Output Generation

   • Generates:
     • SQL DDL scripts
     • ERD (Mermaid format)
     • Data dictionary
     • Load plan with dependency graph

9. Load Plan Engine

   • Computes optimal table insertion order
   • Detects circular dependencies
   • Suggests batching strategy

10. Combine & Explain

• Merges all outputs
• Optional AI explanation of schema decisions

1. Response Output

• Returns structured JSON via webhook:
  • SQL schema
  • ERD summary
  • Data dictionary
  • Load plan
  • Optional explanation

Setup Instructions

1. Activate the workflow and copy the webhook URL
2. Send a POST request with a CSV or XLSX file
3. Configure OpenAI credentials (used by AI agent)
4. Adjust thresholds if needed (FK overlap, retries, confidence)
5. Execute workflow and review generated outputs

Use Cases

• Auto-generate database schema from CSV/Excel files
• Data migration and onboarding pipelines
• Rapid database prototyping
• Reverse engineering datasets
• AI-assisted data modeling

Requirements

• n8n (latest version recommended)
• OpenAI API credentials
• LangChain nodes enabled
• CSV or XLSX input file