Migrations and Seeders

Production database work is an explicit release operation. Lambda startup does not migrate or seed, and source deployment never implies data mutation.

Migration sets

Every migration root has a .minco-migrations.toml sidecar that records a stable set ID, owner, backend, history table, verification tables, and per-version risk.

schema = 1
id = "orders-postgres"
owner = "orders"
backend = "postgres"
history_table = "_minco_migrations"
verification_tables = ["orders", "order_lines"]

[[migrations]]
version = 1
name = "create_orders"
risk = "additive"
path = "0001_create_orders.sql"

The plan hashes the ordered files and metadata. An edited SQL file produces a new digest and invalidates the previous approval.

Inspect without mutation

cargo minco db plan --set orders-postgres --json
cargo minco db status --json
cargo minco db verify --json

Planning is offline. Status and verification are read-only database operations when a connection is explicitly supplied.

Apply an exact migration plan

cargo minco db plan \
  --set orders-postgres \
  --environment production \
  --json

After reviewing the target, history, ordered steps, risks, and digest, pass that exact digest to the mutating command:

cargo minco db migrate \
  --set orders-postgres \
  --environment production \
  --approve-digest REVIEWED_DIGEST \
  --json

The runner records durable receipts. Target/environment mismatches, changed files, missing approvals, or incompatible history fail before mutation.

Classify seed data

Seed sets declare whether data is reference, demo, test fixture, or another reviewed class, plus the allowed environments and preservation behavior.

cargo minco db seed \
  --profile demo \
  --environment local \
  --dry-run \
  --json

A mutating run requires the classification and environment gates to pass. Use verification as a separate read-only stage:

cargo minco db seed --verify --json

Choose an adapter

• sqlx-postgres supports bounded pools for PostgreSQL connection models such as Neon, self-hosted PostgreSQL, RDS, and Aurora; the deployment profile owns connection and cost assumptions.
• sqlx-sqlite supports local, desktop, and persistent single-process profiles with explicit durability limits.
• DynamoDB needs access-pattern-specific ports and adapters. Minco does not emulate relational SQL semantics over it.

Use real-engine behavioral tests for adapter claims. Compiler success alone is not transaction, locking, migration, or provider evidence.