@alswl: 最近一直在找 Agent Dynamic Workflow 的替代方案。 一开始是沿着 Temporal 这条线找的。Temporal 很强,但如果只是想把 agent、脚本、数据处理、运维任务动态串起来,有时候会觉得整套体系有点重。 然…
摘要
作者介绍了 Dagu 作为 Temporal 的轻量级替代方案,用于动态工作流编排,并附带 Northflank 平台的推广内容。
查看缓存全文
缓存时间: 2026/07/02 00:15
最近一直在找 Agent Dynamic Workflow 的替代方案。
一开始是沿着 Temporal 这条线找的。Temporal 很强,但如果只是想把 agent、脚本、数据处理、运维任务动态串起来,有时候会觉得整套体系有点重。
然后翻到了 Dagu,感觉方向很对。
单个 binary 就能跑,流程用 YAML 写,本地文件管理,带 Web UI,也不用额外起 DB 或 broker。
内置动作也挺多:shell、Docker、K8s Job、SSH 这些都有。比较惊喜的是还有 http://harness.run,可以直接把外部 coding agent CLI 接进流程里。
我喜欢它的点是:流程本身就是文件,状态、日志、重试、依赖和界面都帮你兜住了。
对小团队、私有环境、个人自动化,还有 agent workflow 来说,Dagu 这种本地优先的路线反而很舒服。
http://github.com/dagucloud/dagu
Northflank — Deploy any project in seconds, in our cloud or yours.
Source: https://northflank.com/ Run sandboxes, AI models, and databases
The deployment platform for serious workloads
Ship AI, services, databases, and jobs, securely. From commit to production and beyond. In your VPC, or ours.
Northflank Platform
Runs on
Northflank Cloud
or
Microsoft Azure
Amazon Web Services
Google Cloud Platform
Bring Your Own Cloud
Trusted by 2,000+start-upsandenterprises
With you at every step
The operating system for your engineering team
Run containers on command
- Run your most critical services, databases, jobs, inference, and agents
- Spin up ephemeral and preview environments from pull requests
- All your production infrastructure
- Connect Git to trigger builds and deploy
- Host workloads on our cloud or yours
We’ve got you covered on all flanks
If you’re building with AI…
- Deploy and scale OSS models like Llama4, Deepseek
- Secure code execution for code-gen
- Long-running agents (MCP, RAG, TTS, image, video, voice)
- VectorDBs like pgvector
- Spot instances
- Import your GPU K8s clusters
- Jupyter notebooks
- Custom autoscaling
- Multi-cloud support
- Fast multi-read-write storage
...you still need these.
- Secrets and config management
- Deploy applications, APIs, databases, and scheduled jobs
- Build and release pipelines
- Preview, staging, and production environments
- Continuous Integration and Continuous Deployment (CI/CD)
- Templates
- Secure multi-tenancy
- Run in your VPC
Works with any stack
Features
Developer experience
Collapse details
- Choice of UI, CLI, APIs & GitOps
- Runs on AWS, GCP, Azure, Oracle
- Build re-useable templates
Polyglot platform
Expand details
- Run any language or framework
- All envs, from preview to production
- Works with GitHub, GitLab and Bitbucket
Run your AI
Expand details
- Scale GPUs across clouds
- Run untrusted code at scale
- Serve your inference & models
- Support fractional GPU workloads
We fixed Kubernetes
Expand details
- Kubernetes-ready app platform
- Run anywhere, on any Kubernetes cluster, on any cloud
- EKS, GKE, AKS, Rancher, OpenShift, Tanzu ready
End-to-end
A golden path to production
Accelerate every step from push to production with highly configurable self-service workflows, pipelines, templates, and GitOps.
Securely deploy preview, staging, and production environments with observability tooling, backups, restores, and rollbacks included.
Learn moreabout Northflank features
Testimonials fromWeights and Polyguard
From the front lines
Northflank helped us scale to over 3M users with an engineering team of two. It allows us to run 10,000 AI training jobs and half a million inference runs per day, without us having to worry about autoscaling or spot instance orchestration.We don’t waste time or money on infrastructure, so we can focus on building product.

Founder @ Weights
We run it better
Flexibility without compromise
Northflank seamlessly integrates with your preferred tooling, and can accommodate any tech stack.
You get the same exceptional developer experience, and total control over your data residency, deployment regions, security, and cloud expenses.
Learn moreabout the Northflank platform
Testimonials fromSentry and Writer
Endorsed by infra veterans
Northflank is way easier than gluing a bunch of tools together to spin up apps and databases. It’s the ideal platform to deploy containers in our cloud account, avoiding the brain damage of big cloud and Kubernetes. It’s more powerful and flexible than traditional PaaS – all within our VPC.Northflank has become a go-to way to deploy workloads at Sentry.

Co-Founder and CPO @ Sentry
K8s should work for you
Kubernetes, unleashed
Northflank leverages Kubernetes as an operating system to give you the best of cloud native, without the overhead.
Deploy to Northflank’s cloud for maximum simplicity, or connect your GKE, EKS, AKS, or bare-metal to deliver a managed platform experience in minutes.
Learn moreabout Northflank in Your Cloud

Built for scale
You’re in good company
Since 2019, teams have used Northflank to run everything from enterprise products to high-scale AI infrastructure. Whether it’s one container or one thousand, Northflank holds the line.
80k+
Developers in production
Run anything. Anywhere.
Deploy in your preferred cloud
In the strict industry of pharmaceutical manufacturing, precision, accuracy, and compliance are non-negotiable. Northflank has empowered our team to maintain complete accountability for every aspect of our software development and deployment, providing us with clear change management, release control and traceability features. The additional flexibility of the platform to make use of Compliant Cloud via BYOC and optional on-prem deployment means that we can meet our customers wherever they may need us.

Bartlomiej Baran
Co-Founder and CTO @ Blynksolve
Making Kubernetes boring
相似文章
@realWeZZard: https://x.com/realWeZZard/status/2062105579649380748
开发者@realWeZZard分享了他在Anthropic推出Dynamic Workflow四个月前自行构建charge插件的经历,详细讨论了subagent编排的设计取舍、成本控制和最终选择amplify的原因,并提供了实用决策树。
@RealCodedAlpha: 这篇关于 Claude Code Dynamic Workflows 的 9 步指南真的讲透了! 很多人玩 multi-agent 都是直接开 swarm,结果一堆冲突、低质输出和 merge 地狱。 作者把核心说清楚了:结构化的 loo…
这篇推文介绍了Claude Code Dynamic Workflows的9步指南,强调了结构化循环和多智能体工作流的最佳实践,包括人工审查、工作树隔离和自动返工,指出这是将agent swarm从玩具变成生产力的关键。
@vintcessun: 最近不少人在折腾agentic workflow,prompt调了半天还是容易跑偏。其实问题常在runtime层:循环没预算、工具权限太宽、压缩丢了状态。DenisSergeevitch这个agents-best-practices sk…
讨论agentic workflow中常见的runtime问题(循环预算、工具权限、压缩状态丢失),推荐DenisSergeevitch的agents-best-practices资源,提供provider-neutral的参考,强调将权限、预算、观测作为显式机制。
@chasen_liao: https://x.com/chasen_liao/status/2077219202608545835
本文探讨了提示词工程升级为Agent工程的趋势,强调通过AGENTS.md等结构化上下文管理AI代理,并分享了一套最小闭环的工作流方法论。
@yanhua1010: 目前看到关于 “Agentic Engineering Workflow”的最完整的介绍 花了一个小时完整看完了,完全可以做成一个付费教程。 内容涵盖了tmux,agent记忆,skills,语音输入,长任务执行,并行worktree管理…
推荐了一份关于“Agentic Engineering Workflow”的完整介绍,涵盖了tmux、agent记忆、技能、语音输入、长任务执行、并行worktree管理、多agent调度以及可视化HTML编辑器Lavish和代码变更校验流水线no-mistakes。