Stelvio: Serverless AWS for Python Devs | Hacker News Reader

Stelvio: Serverless AWS for Python Devs

github.com

48 pointsmilsebg26 comments

Hello,

author of Stelvio here. Stelvio is in very early alpha far from any production use, but it's able to deploy simple API gateway, lambda and dynamo db tables with just few lines of code.

Idea is that Python devs should be able to deal with infra with Python only and minimal boilerplate - simple thing should be simple.

I'm currently working on v 0.2.0 which will have support for lambda dependencies & layers, Dynamo indexes and CORS and authorizers for API gateway.

I hope to release that in April.

For v0.3.0, stelvio will get it's own CLI too.

Any feedback is welcomed here or michal at stelvio.dev

Michal

Hi! Honest question, why did decide on using pulomi instead of just using the boto3 Python library to interact with aws? Feels like it adds a lot of overhead where something like boto3 could just be abstracted away?

Hi! Good question, few reasons:

1. state management & stages - pulumi already deals with state and has support for stages (stacks) which stelvio will need.

2. there are things outside of AWS that stelvio might support later (cloudflare, etc.)

Initially I was gonna use AWS CDK which I use at work but Pulumi seems to have faster deployment and option to use other things than AWS. (Although AWS is main focus of Stelvio at leats for now)

Why not Zappa?

https://github.com/zappa/Zappa

Hi! AFAIK in Zappa you don't define infra in Python but rather in json. Also the vision of Stelvio is broader than just serverless. I wrote about my vision here https://blog.stelvio.dev/why-i-am-building-stelvio/ if you're interested.

The cloud is too complex.

While it's good to enable proprietary solutions like AWS, we should always make sure what we make is not bound to it.

This might mean that stelvio becomes a part of something more general.

Not trying to detract from the OP, but to provide a combination I have running in production for 2 years.

- aws cdk (with outputs) - https://github.com/zappa/Zappa - a python script which stitches outputs from cdk into the zappa config - extra python scripts to do a few small things post deployment. Zappa has some bugs that would be tedious to fix vs 100 lines of python.

Our product has the luxury of only being used during stock market hours and really thrives in serverless everything. We use rds serverless (v2) along with lambdas.

Some of our work is heavy and we’ll dynamically spin up an ecs container which has the hallmarks of a normal django app: redis + celery queues. We try to saturate the ecs container resources with this type of setup. After the container is done, it’ll shutdown.

I was super skeptical of this 2 years ago. 4 envs costing ~$2k/month. I would do this setup again if the product warrants windowed usage.

This is great. Just this morning I was just lamenting how I hate creating AWS API Gateways, and I wish I could do it in python like I do with FastAPI.

This looks to be exactly what I was looking for. When you are production ready this is something I'll use.

I recently made a serverless lambda/dynamo webapp following similar patterns and wrote about the experience here: https://johnscolaro.xyz/blog/serverless-python-websites

I think it's amazing the python community has like 5 half baked solutions to this problem, all of which are either abandoned, poorly monetized, or have a janky UI. I mean we have: Zappa, Chalice, Serverless, and if you attempt to do it yourself, do you use Cloudformation, CDK, or AWS SAM?

Following with interest, I think there is room for better tool in this space.

Nice writing and summary there. As an author of stelvio I felt the same. It's not that there're no tools. But none of them is covering enough AWS surface while being truly focused on Python devs. Some are general infra tools (CDK, Pulumi) focused on infra people, where python is not primary language, some are Python tools like Chalice (likely abandonded) and some are meant for Python devs but use json (Zappa).

I wrote my own thinking here https://blog.stelvio.dev/why-i-am-building-stelvio/

I 'm working hard on Stelvio hoping it can become the best tool to do infra in Python.

Btw. If you'd be willing to share more details about what you want/miss in python cloud tooling could you drop me an email at michal at stelvio.dev ? I'm happy to talk. Applies to other interested people too!

Surprised to not see a comparison Serverless Framework: https://www.serverless.com/

The above recently did a rug pull and being developed in JS means it's not great for Python devs to maintain an open source version, so good to see alternatives in Python.

There is also Mangum [0], which can wrap up any ASGI app and deploy it as a lambda.

[0] https://mangum.fastapiexpert.com/

There's a maintained drop-in replacement called osls: https://github.com/oss-serverless/serverless

Eugh. What rug pull do you mean?

Serverless Framework changed their license terms in v4 to a paid model. It’s also not getting a lot of new development or community support anymore. It’s slowly dying and time to move off of it.

Have a look at https://github.com/oss-serverless/serverless

Neat tool!

Not sure I agree with the choices on the AWS side, particularly lambda and dynamo. They seem great when you first start, but I've found the pain comes with maturing code.

As professionals, we should always plan for our code to live longer than we thought originally, and supporting new use cases not in the initial design.

A cheap always on server and a hosted RDBMS are going to go a lot further for next to the same mental overhead.

> Neat tool!

thanks! (author here not OP)

lambda/api/dynamo is just a start. I have very clear vision where i want to take this and it's not limited to lambda/api/dynamo. but these are easier to start with before you get to VPC, NAT, security groups etc. etc.

Would love something that sets up EKS and ingress without me tearing my hair out and actually wishing I were using Azure (yikes).

Both of these tools hit the sweet spot starting out—good performance for cheap (not entirely true if you consider the free tier of sql services, but I've had bad experiences with those). It's the lock-in and traffic that kills you.

How is this better than AWS CDK, and why are you not using CloudFormation? Seems like the sane thing to do in 2025

I guess this question is not aimed at me as I'm not OP but I'll try to answer as the author of stelvio. I use CDK at my day job every day.

I like CDK, it was huge step forward compared to Cloud Formation, Terraform etc.

But I think infra tools can be still made better for developers. In stelvio you can do this:

  table = DynamoTable(
      name="todos",
      fields={
          "username": AttributeType.STRING,
          "created": AttributeType.STRING,
      },
      partition_key="username",
      sort_key='created'
  )

  api = Api("todo-api")
  api.route("POST", "/todos", handler="functions/todos.post", links=[table])
  api.route("GET", "/todos/{username}", handler="functions/todos.get")

It creates:

- DynamoDB table

- Lambda function with IAM (policy - with permissions for dynamo table, role)

- API Gateway (resources, methods, integrations, role, stage & deployment)

- CloudWatch log groups

It would be much more code in CDK.

Stelvio aims to make common things much simpler while allowing you to change details if you need to.

CDK has high level constructs like aws_ecs_patterns.ApplicationLoadBalancedFargateService etc. It seems like the above could be implemented as high level CDK constructs. So you could make common things simpler, but have the full CDK to change details if needed.

How does this differ from Chalice? Or is it a replacement now that it is being deprecated?

Hey, I was looking for similar tool before starting building stelvio and I found Chalice. It looked good, it's a nice tool but:

- as you mentioned, seems to be abandoned

- chalice mixes infra code with app code within the file (decorators), stelvio keeps them separate

- in chalice you configure lambdas using json. I strongly believe that everything should be done in python

- chalice is limited to lambda/api/queues, stelvio aims to go further/broader

hope this explains, happy to answer any other questions

This also seems a bit like AWS Chalice reinvented, which might be a good thing as AWS has silently abandoned it.

Reply on news.ycombinator.com