Ledger-first write path
Command, aggregate, append, and SyncInTxn projection share one database transaction. Balances read their own writes after commit.
Open source · Go 1.24+ · MIT
Event-sourced ledgers,
built for Go finance stacks
go-sequent is a CQRS and event sourcing library for payments, wallets, and ledger workloads. It keeps the write path boring and correct: one transaction, context-scoped UnitOfWork, SyncInTxn projections.
go get github.com/go-sequent/go-sequent@latestWhy teams reach for it
Inspired by Sequent’s domain loop, implemented as idiomatic Go. No Ruby database import, no web framework lock-in.
Integrity without ceremony
Per-stream hash chains detect tampering. Optimistic concurrency and tenant-scoped unique keys protect concurrent appends.
Finance defaults on
Decimal money, command audit metadata, idempotency keys, AuthZ hooks, and shared-schema tenancy are first-class.
Pure library shape
No bundled HTTP server. Embed go-sequent in your process, wire PostgreSQL or memory stores, and keep ops in the CLI.
The core loop
01
Command
Capture intent with typed commands and finance-oriented Meta.
02
Aggregate
Decide and Apply events. State changes only in appliers.
03
EventStore
Append with OCC, unique keys, and a per-aggregate hash chain.
04
Projector
Update ledger views inside the same write transaction.
Locked defaults for finance
Product decisions stay explicit so operators and auditors can reason about the system without hunting folklore.
- Ledger projections run SyncInTxn
- Money is decimal — never float64
- Integrity via hash chain; no event signatures
- Shared schema +
tenant_id - GDPR path: tombstone only
- Outbox via PostgreSQL table polling
demo, err := bank.NewDemoApp(bank.DemoOptions{
StoreFactory: memoryFactory,
Idempotency: idempotency.NewMemory,
Authorizer: bank.AllowBankCommands,
EnableTenancyMW: true,
})
ctx = tenancy.WithTenant(ctx, tenantID)
err = demo.App.Service.Execute(ctx, &bank.OpenAccount{ /* … */ })
bal, _ := demo.Projector.Balance(accountID)Start with the bank example
In-memory demo for always-on tests; Postgres sample for production-shaped balances.