github.devotd.
NewDeveloper tools

dbos-inc/dbos-transact-py 3.3.0 — Postgres-backed durable workflows

dbos-inc/dbos-transact-py is a Python durable workflow library that checkpoints execution state in Postgres and resumes from the last completed step after a failure; the latest release is 3.3.0.

Repositorydbos-inc/dbos-transact-py ↗

dbos-inc/dbos-transact-py is a Python durable workflow library that checkpoints execution state in Postgres and resumes from the last completed step after a failure; the latest release is 3.3.0.

GitHub description: Database-Backed Durable Python Workflows

stars
★ 1.6K
Forks
108
Language
Python
License
MIT
Latest release
3.3.0
Read original source ↗View on GitHub ↗

dbos-inc/dbos-transact-py is a Python workflow library that records execution state in Postgres, so a run that breaks midway continues from the last completed step. The repository describes itself as 'Database-Backed Durable Python Workflows'. A durable workflow is an execution flow that keeps its progress even when it fails. The latest release is 3.3.0, and the given material contains no changelog for this version, so what changed cannot be confirmed.

Marking a function as a workflow with step markers checkpoints state in Postgres and resumes automatically after a failure. It offers a durable queue, exactly-once event handling, durable scheduling and notifications, and workflow management driven by code, with no separate orchestrator or message broker. A durable queue is a waiting line for jobs that does not lose state across restarts. The license is MIT, the primary language is Python, the repository has 1642 stars, and it was created on July 19, 2024.

Still, this record carries no performance figures and no comparison with other orchestration tools. The only evaluation score left in the record is momentum 2.1, and the novelty, technical and quality items are unfilled. How fast checkpointing and resumption are in a real service, and under which conditions exactly-once event handling holds, are not written down in the documentation. The range of workflows this library can handle can only be known through its own measurements.