github.devotd.
NewDeveloper tools

hatchet-dev/hatchet v0.110.5 — Background Job and AI Agent Orchestration

Covers hatchet-dev/hatchet's latest release v0.110.5, the queuing, automatic retry, and durability features of its Go-based orchestration engine, and the limitation that release notes could not be confirmed.

Repositoryhatchet-dev/hatchet ↗

Covers hatchet-dev/hatchet's latest release v0.110.5, the queuing, automatic retry, and durability features of its Go-based orchestration engine, and the limitation that release notes could not be confirmed.

GitHub description: hatchet-dev/hatchet

stars
★ 8.2K
Forks
0
Language
Go
Read original source ↗View on GitHub ↗

The latest release tag of hatchet-dev/hatchet is v0.110.5. Hatchet is an orchestration engine implemented in Go that coordinates background jobs, AI agents, and durable (durability) workflows at scale.

The README states that Hatchet is a platform that orchestrates background jobs, AI agents, and durable workflows for Python, TypeScript, Go, and Ruby applications at scale. It brings queues and job dependencies, retries, and observability (real-time monitoring) together in one place to control the flow of execution. It offers job enqueueing (queuing), automatic retry of failed jobs, and durability that carries a workflow forward even when a process breaks midway, all at once. Notifications and logging are used to check status, and it can be used as the Hatchet Cloud service or self-hosted. It is a structure that consolidates queue processing, retries, and status checks — normally managed separately when running background jobs — into a single platform. The implementation language is Go, the license is MIT, and the repository was created on December 15, 2023. It currently has 8158 stars.

However, the parts that can serve as evidence are limited. No release notes explaining what changed in v0.110.5 could be confirmed, so only the version tag is visible in this release. In the evaluation score items, only momentum is filled in at 4.6, while novelty, technology, and quality are empty. There is also no comparison baseline (benchmark) such as workflow throughput, so it is too early to judge whether performance has improved. Self-hosting is possible under the MIT license, but the operating environment must be prepared by yourself. Given that only the feature scope and the tag can be confirmed right now, it is safer to decide on adoption after the release notes are filled in.