marpy.io
专为Python栈打造的AI编程平台,提供原生代码补全、重构和AI生成模块,让开发者快速从想法到部署应用。
Maker 说
嘿哟!
过去几年我眼看着朋友和客户用 vibe-code 搓出几百个应用。几乎每个项目最后都在跟 JavaScript 后端死磕,而用 FastAPI 写 Python 代码不过二十行。所以为了根治这痛点,我一怒之下搞出了 marpy。
marpy 是什么?
它是一个基于浏览器的 IDE、部署平台和 AI 助手,纯为 Python 生态打造的(Flask、FastAPI、Django)。想象一下 Replit 或 v0,但专为我们 Python 开发者而生。
为什么是 Python?
我开玩笑说:“如果你大声念你的 Python 代码,听起来像个山顶洞人,那就写对了。”但说真的,语法简单,特别适合 LLM。
技术栈是什么?
核心是 Python、MariaDB、Redis、Docker 和 K8s。内置 git 集成,万一 AI 真的接管互联网了,你的代码也还在你手里。
我的理念
我相信人工智能是针对特定问题的高度专业化解决方案,不是银弹;我也不喜欢“AI 能解决一切问题”的炒作。
我还用 marpy.io 构建了 marpy.io。我每天都在吃自己的狗粮,用我自己的 Flask/MariaDB 栈,确保“刚刚好”原则真能帮人交付能用的产品。
欢迎你加入我。
——Sethers
I've spent the last few years watching my friends and clients vibe-code hundreds of apps into existence. Almost every app has ended with us fighting JavaScript back-ends that would be twenty lines of FastAPI Python code. So, to deal with the pain I rage-built marpy.
What is marpy?
It’s a browser-based IDE, deployment platform, and AI assistant built exclusively for the Python ecosystem (Flask, FastAPI, Django). Think Replit or v0, but for us Python devs.
Why Python?
I joke that "if you read your Python code out loud and you sound like a caveman, you did it right." But, really, the syntax is simple and lends itself really well to LLMs.
What's the stack?
Python, MariaDB, Redis, Docker, and K8s at the core. With git integrations built in so when AI finally takes over the internet, you'll still have all of your code.
My Philosophy
I believe Artificial Intelligence is a highly specialized solution for specific problems, not a silver bullet; and I don't like the "AI will solve all the problems" hype.
I also built marpy.io on marpy.io. I'm dogfooding my own Flask/MariaDB stack every day to make sure the "Just Enough" principle actually helps ship products that work.
It would be awesome for you to join me.
-Sethers
我很好奇,当你的想法涉及有状态的东西——比如一个 Celery beat 任务,一个需要数据库连接的后台 worker——你们是如何处理“从想法到部署应用”这个鸿沟的?这正是我看到 vibe-coding 工具崩溃的地方:生成的代码在本地能跑,但它假设了一个在生产环境中并不存在的环境。Marby 的上下文模型是尽早捕捉到基础设施假设,还是等到部署时才暴露出来?
感觉现在更多 AI 工具是通过深入细分领域而不是追求广度来获胜的。