Honker: SQLite Extension for NOTIFY/LISTEN, Task Queues, and Cron in One .db File

Honker is a SQLite loadable extension plus a multi-language binding suite that turns any `.db` file into a complete background-jobs and pub/sub platform. It exposes cross-process NOTIFY/LISTEN primitives modelled on Postgres, durable work queues with retries and delayed jobs, append-only event streams with per-consumer offsets, named locks, rate limiting, and a leader-elected cron scheduler — all as rows in the same database file as application data, committed atomically with business writes. Language bindings ship for Python, Node.js, Bun, Rust, Go, Ruby, Elixir, and C++. The library is ORM-compatible, so the queue can share connections with SQLAlchemy, Django ORM, Drizzle, Kysely, sqlx, GORM, ActiveRecord, or Ecto. The core extension is written in Rust; the broader codebase is mixed Python and Rust. Claimed performance (per project documentation): cross-process wake latency around 0.7 ms p50 with 1–2 ms median on Apple Silicon, throughput of "thousands of messages per second on a modern laptop," and roughly 0.35% of one CPU core consumed by the polling thread at 1 kHz. Durability guarantees inherit from SQLite ACID: a rolled-back transaction drops any enqueued jobs along with business writes, SIGKILL mid-transaction is safe, and worker crashes are recovered via a `visibility_timeout_s` claim that defaults to 300 seconds. The project is explicitly experimental — the author warns that the API may change, the npm package lives under a personal scope, and the design is optimized for the single-`.db`-file case. Deliberately out of scope are Celery-style task pipelines, chains, groups, chords, multi-writer replication, and DAG-based workflow orchestration. The pitch targets the operational simplicity of SQLite-primary deployments: rather than adding Redis plus a broker just to run background jobs, the queue lives in the same file as the application data, with no second datastore to back up or operate. See PRAGMA data_version: SQLite's Global Commit Counter as a Cross-Process Polling Primitive for the implementation trick that makes sub-millisecond cross-process notification possible without a daemon.