@iterativeflow/sqlite

A SQLite backend for iterativeflow — the four ports (Store / Queue / Timer / Wakeup) over one SQLite database. Embedded and edge: a local file, Turso/libSQL, or a Cloudflare Durable Object's SQLite storage, all through the same Sql seam.

State and dispatch share the database, so the transactional outbox commits in one BEGIN…COMMIT — the same atomic model as the Postgres backend, no Lua or two-phase choreography. Passes the same conformance suites as the Postgres, DynamoDB, and Redis backends.

import { createClient } from "@libsql/client";
import { createSqliteBackend, applySchema, libsqlDb } from "@iterativeflow/sqlite";
import { createEngine } from "@iterativeflow/core";

const sql = libsqlDb(createClient({ url: "file:workflow.db" })); // or a Turso url
await applySchema(sql);
const engine = createEngine(createSqliteBackend(sql), [
  /* your flows */
]);

Notes

• @libsql/client is a peer dependency — bring your own client (a local file, :memory: won't share across transactions, or a Turso remote). Any adapter satisfying the two-method Sql interface works, so a Durable Object's ctx.storage.sql can drive the same backend.
• Wakeup is in-process (like DynamoDB/Redis); a run completing in another process is caught by the poll-first result().
• Single-writer: SQLite serializes writes, so claim runs a plain SELECT+UPDATE in a transaction — no SKIP LOCKED needed.
• Run applySchema(sql, prefix?) once before use.