keepsake

Let it be forgotten, as a flower is forgotten, Forgotten as a fire that once was singing gold.

— Sara Teasdale, "Let It Be Forgotten" (1920)

keepsake stores relations that a subject holds until policy ends them: a trusted tag, a 24-hour mute, an entitlement, a hold, a risk flag, a feature gate. Writes are idempotent, expiry runs on a schedule you set, state is queryable, and apply/revoke produce typed audit records.

The core crate is persistence-agnostic and synchronous. The keepsake-sqlx adapter stores state through SQLx with migrations and query helpers. Postgres is the default backend; SQLite and MySQL are available behind feature flags.

Where it fits

Use the crate directly for a Rust service backed by Postgres, SQLite, or MySQL. For other stacks, the schema, indexes, and lifecycle rules are documented so you can port them to another language, framework, or database.

Some responsibilities stay with your application. Keepsake does not join your entity tables, make authorization decisions, invalidate distributed caches, or consume domain events. It stores relation state and expiry; authorization reads those relations later.

Install

cargo add keepsake keepsake-sqlx
cargo add sqlx --features postgres,runtime-tokio,tls-rustls

Run the embedded migration with a sqlx::PgPool:

use keepsake_sqlx::KeepsakeRepository;
use sqlx::PgPool;

let pool = PgPool::connect(&database_url).await?;
let repo = KeepsakeRepository::new(pool);
repo.migrate().await?;

For SQLite or MySQL, disable default features and enable the target backend, then construct SqliteKeepsakeRepository or MySqlKeepsakeRepository with the matching SQLx pool. The repository type and pool type are coupled at compile time, and migrations also record a backend marker so a schema initialized for one driver is not silently reused by another.

Operations

Migrations: keepsake-sqlx embeds SQLx migrations. Run them at startup or from your migration runner. Disable the migrations feature to vendor the SQL into another framework.
Audit: apply and revoke take command objects and record actor/context metadata with each lifecycle change. They are the canonical mutation path.

Lifecycle semantics are always on. Idempotency, duplicate-active prevention, deterministic ordering, opaque subjects, and indexed read paths are part of the contract. An optional relation-definition cache is available; caching active state is left to the application.

Why it exists

I'd written this pattern ad-hoc across production services in compliance-heavy domains, where auditability and determinism are requirements. keepsake is the consolidated version, so you pull in one implementation instead of re-deriving the same rules in every project.

License

Licensed under either of:

Apache License, Version 2.0
MIT license