SeedForge

Database seeding tool. Connects to your DB, reads the schema, generates realistic data with valid foreign keys.

pip install seedforge
seedforge connect postgresql://user:pass@localhost/mydb
seedforge generate --rows 1000

Why

I got tired of writing seed scripts by hand every time I start a new project. You know the drill — empty dashboards, nothing to test against, and if you copy prod data you're asking for GDPR trouble.

SeedForge reads your schema and figures out the rest. It knows that orders.user_id points to users.id, so it fills users first. It knows that a column called email should look like an email, not random gibberish.

Install

pip install seedforge          # PostgreSQL + SQLite
pip install seedforge[mysql]   # + MySQL/MariaDB
pip install seedforge[mssql]   # + SQL Server
pip install seedforge[ai]      # + AI providers
pip install seedforge[all]     # everything

Usage

# save connection (writes .seedforge.yaml)
seedforge connect postgresql://user:pass@localhost:5432/mydb
seedforge connect mysql://user:pass@localhost:3306/mydb
seedforge connect sqlite:///path/to/db.sqlite
seedforge connect mssql://sa:pass@localhost/mydb

# see what's in the database
seedforge inspect

# generate and insert
seedforge generate --rows 100

# just preview, don't touch the DB
seedforge generate --rows 10 --dry-run

# export instead of inserting
seedforge generate --rows 1000 --export sql
seedforge generate --rows 1000 --export json

# same data every time
seedforge generate --rows 100 --seed 42

# specific tables only (pulls in FK parents automatically)
seedforge generate --tables orders,payments --rows 50

# wipe tables first
seedforge generate --rows 100 --clean

AI mode

If you want smarter data (realistic product names, proper bios, etc.), you can plug in an API key. SeedForge auto-detects the provider:

seedforge ai-generate --api-key sk-ant-...   # Anthropic
seedforge ai-generate --api-key sk-...       # OpenAI
seedforge ai-generate --api-key AIza...      # Google Gemini
seedforge ai-generate --api-key gsk_...      # Groq

Or set an env var (ANTHROPIC_API_KEY, OPENAI_API_KEY, GEMINI_API_KEY, GROQ_API_KEY) and just run seedforge ai-generate.

Only schema metadata goes to the API — table and column names, never your actual data.

How it works

1. Connects to your database and reads information_schema (PostgreSQL/MySQL) or PRAGMA (SQLite)
2. Builds a dependency graph from foreign keys, topological sort gives the insertion order
3. For each column, picks a generator based on the name — email gets a realistic email, price gets a decimal, created_at gets a recent timestamp, and so on (80+ patterns)
4. Foreign key columns get real IDs from already-generated parent rows
5. Batch inserts everything in a single transaction

What it recognizes

Some examples — there are 80+ patterns total:

It's also context-aware: name in a users table gives person names, in organizations — company names, in products — product names.

Config

seedforge connect creates a .seedforge.yaml:

db_url: postgresql://user:pass@localhost:5432/mydb
default_rows: 100
default_schema: public
seed: 42
exclude_tables:
  - _prisma_migrations
  - django_migrations

Migration tables (_prisma_migrations, django_migrations, alembic_version, etc.) are excluded automatically.

Databases

• PostgreSQL (+ CockroachDB)
• MySQL / MariaDB
• SQLite
• Microsoft SQL Server

Contributing

PRs welcome. See CONTRIBUTING.md.

License

MIT