Show HN: Stepped Actions – distributed workflow orchestration for Rails(github.com)
github.com
Show HN: Stepped Actions – distributed workflow orchestration for Rails
https://github.com/envirobly/stepped
Stepped is a Rails engine, extracted out of Envirobly where it powers tasks like application deployment, that involve complex, out-of-the-band tasks like DNS provisioning, retries, waiting for instances to boot, running health checks and all the fun stuff of a highly distributed networked system.
15 comments
Nice work. Thanks for sharing it! I've been thinking about using something like this for LLM agent workflows - the outbound action pattern would work well for tool calls that need to wait on external APIs.
I'm working on DSPy.rb [1] and this could pair nicely for multi-step reasoning chains.
Curious - any plans for async gem support?
[1] https://oss.vicente.services/dspy.rb/
I'm working on DSPy.rb [1] and this could pair nicely for multi-step reasoning chains.
Curious - any plans for async gem support?
[1] https://oss.vicente.services/dspy.rb/
Not sure how async gem would fit into this. The backbone of Stepped Actions is ActiveJob which handles execution. Where do you think this would fit in?
You’re right they’re different models. The path would be Falcon + async-job-adapter-active_job. I am exploring eliminating worker processes, so jobs run as fibers in the same process, yielding during I/O.
How does DSPy.rb differ from BAML?
Both model prompts as functions. BAML is a DSL - write .baml files, generate code, get validated structured outputs.
DSPy is a programming paradigm. I like to look at it like the MVC for the Web. You define Signatures[0]: typed contracts governing the relationship between your models and your app. Signatures model prompts as functions too, without leaving Ruby. Then compose them into modules (Predict, ChainOfThought, ReAct, your own). The framework can automatically optimize prompts based on your metrics.
DSPy.rb brings the DSPy paradigm's tooling (optimizers, evaluation loops) to Ruby. Comes with OpenTelemetry OOTB. It also borrows BAML's schema format for 85% token savings vs JSON Schema in complex signatures. [1]
Everyone is talking about prompt, context, and harness engineering -and I agree they are good ways to frame how to build workflows and agents- this is just programming really.
[0] https://oss.vicente.services/dspy.rb/core-concepts/signature...
[1] https://oss.vicente.services/dspy.rb/articles/baml-schema-fo...
DSPy is a programming paradigm. I like to look at it like the MVC for the Web. You define Signatures[0]: typed contracts governing the relationship between your models and your app. Signatures model prompts as functions too, without leaving Ruby. Then compose them into modules (Predict, ChainOfThought, ReAct, your own). The framework can automatically optimize prompts based on your metrics.
DSPy.rb brings the DSPy paradigm's tooling (optimizers, evaluation loops) to Ruby. Comes with OpenTelemetry OOTB. It also borrows BAML's schema format for 85% token savings vs JSON Schema in complex signatures. [1]
Everyone is talking about prompt, context, and harness engineering -and I agree they are good ways to frame how to build workflows and agents- this is just programming really.
[0] https://oss.vicente.services/dspy.rb/core-concepts/signature...
[1] https://oss.vicente.services/dspy.rb/articles/baml-schema-fo...
Congratulations on shipping this, I’m sure folks will find it useful!
The rails native way to do this is to track state in a db row and queuing “next step” jobs as the data changes. This can get verbose especially for smaller pass/fail workflows. However, I find this works better (not worse imo) in more complex workflows as the state is tracked, queryable, can be surfaced in UIs, and resumed “manually” in the event of an outage.
The rails native way to do this is to track state in a db row and queuing “next step” jobs as the data changes. This can get verbose especially for smaller pass/fail workflows. However, I find this works better (not worse imo) in more complex workflows as the state is tracked, queryable, can be surfaced in UIs, and resumed “manually” in the event of an outage.
Thanks :-)
Nice. I have a simple system for typescript [1] where you can string tasks together like:
import { Workflow } from "@workglow/task-graph"; const workflow = new Workflow(); workflow .DownloadModel({ model: ["onnx:Xenova/LaMini-Flan-T5-783M:q8", "Universal Sentence Encoder"], }) .TextEmbedding({ text: "The quick brown fox jumps over the lazy dog.", }); await workflow.run();
It automatically caches results of steps if you say `new Workflow({outputCache});`
PS: the example above uses a local onnx model (it switches on model name) which is a good candidate for caching, but can be anything.
You can play around writing this way if you open the console at the web example [2] which has fun console formatters not enough people know about or use in DevTools. You can write code in the console and the example rewrites it as a json format in the web page (and visa-versa).
Or just use the free web app with local user account and local models to play around with. [3]
[1] https://github.com/workglow-dev/workglow
[2] https://workglow-web.netlify.app/
[3] https://workglow.dev
import { Workflow } from "@workglow/task-graph"; const workflow = new Workflow(); workflow .DownloadModel({ model: ["onnx:Xenova/LaMini-Flan-T5-783M:q8", "Universal Sentence Encoder"], }) .TextEmbedding({ text: "The quick brown fox jumps over the lazy dog.", }); await workflow.run();
It automatically caches results of steps if you say `new Workflow({outputCache});`
PS: the example above uses a local onnx model (it switches on model name) which is a good candidate for caching, but can be anything.
You can play around writing this way if you open the console at the web example [2] which has fun console formatters not enough people know about or use in DevTools. You can write code in the console and the example rewrites it as a json format in the web page (and visa-versa).
Or just use the free web app with local user account and local models to play around with. [3]
[1] https://github.com/workglow-dev/workglow
[2] https://workglow-web.netlify.app/
[3] https://workglow.dev
I have implemented (or managed a team that implemented) this concept four separate times as an internal library somewhere. Each one included slightly different affordances and addressed a few different concerns that stem from the domain that it was built for.
Props for extracting it and offering it up as a library. I'll be interested to compare it to the implementations I've seen, and see what you've added that I've not seen before.
Props for extracting it and offering it up as a library. I'll be interested to compare it to the implementations I've seen, and see what you've added that I've not seen before.
Seems pretty similar to https://github.com/radioactive-labs/chrono_forge which is what I found when I typed in "rails durable execution patterns" into Google. Have you seen this and if so, how do you think it compares?
That looks cool. It's a different approach with different features.
This looks useful. I've been exploring similar durable execution patterns in Go recently to avoid the complexity of Temporal for smaller workflows.
How does stepped_actions handle the state between steps? Does it persist to the DB after every single action to handle crash recovery, or is it more optimistic?
Good luck with the launch
How does stepped_actions handle the state between steps? Does it persist to the DB after every single action to handle crash recovery, or is it more optimistic?
Good luck with the launch
I recently started using this DBOS in Go and it’s been good.
https://github.com/dbos-inc/dbos-transact-golang
I have River on my list to try someday too
https://riverqueue.com/pro
https://github.com/dbos-inc/dbos-transact-golang
I have River on my list to try someday too
https://riverqueue.com/pro
Yes, state is persisted to DB upon every change. Action exceptions are handled gracefully and natural part of the system, they simply fail the action. Crash recovery is build-in thanks to checksums and ActiveJob, if you're using the right adapter, like GoodJob or SolidQueue where crash recovery is guaranteed.
the write up was well done - on why the pivot. & who his actual target market is