我将我的氛围代码技术栈成本优化到了 0 美元。
3 分•作者: udit_50•6 个月前
自从接触到“氛围编码”以来，我一直在尝试构建产品。我做过面向消费者的产品，也做过一些……嗯，昂贵软件的内部克隆。但从一开始我就知道一件事——“氛围栈”很贵。
最初我尝试了很多工具——Bolt、v0、Replit、Lovable 等，其中 Replit 给我的结果最好（是的，由于我选择的应用程序，我可能会有偏见）。但我经常每月要支付 25-200 美元。其他费用，如 API、模型等，使我的月账单超过 300 美元。与雇佣开发人员相比，这是否具有成本效益？是的。但这物有所值吗？不。
因此，几个月来，我优化了我的整个技术栈，使其在内部使用时要么免费（或成本极低），要么在面向消费者的产品上保持更精简的成本。
以下是目前整个技术栈的构成：
IDE - 谷歌的 AntiGravity (100% 免费 + 使用学生证可获得更高权限)
AI 文档 - SuperDocs (100% 免费 & 开源)
数据库 - Supabase (Nano 计划免费，满足基本需求)
身份验证 - Stack Auth (免费，最多支持 1 万用户)
LLM (AI 模型) - OpenRouter 或 Gemini，通过 AI Studio 进行测试，并使用 Unsloth AI 的自定义调优模型进行生产。（您可以使用 Unsloth 在 Google Colab Notebook 中微调模型）
版本维护/分发 - Github/Gitlab (两者完全免费且开源)
更快的部署 - Vercel (免费套餐足够用于业余项目)
分析 - PostHog、Microsoft Clarity & Google Analytics (这三个都是免费的，并且用于不同的跟踪，我建议同时使用它们)
这就是开发者的清单！我知道我可能遗漏了一些东西。如果是这样，请随时问我或在评论中列出来。如果您对特定内容有任何疑问，也可以问我。
查看原文
Since vibe coding came into existence, I have been experimenting with building products a lot. Some of my products were consumer facing and some.. well, internal clones of expensive software. However, since beginning, I knew one big thing - the vibe stack was expensive.<p>I initially tried a lot of tools - Bolt, v0, Replit, Lovable, etc. out of which Replit game me the best results (yes, I can be biased due to my selection of applications). But I often paid anywhere from $25-$200/mo. Other costs like API, models, etc. made monthly bills upward of $300/mo. Was it cost effective when compared to hiring a developer? Yes. Was it value for money? NO.<p>So, over the months, I optimised by complete stack to be either free (or minimal cost) for internal use or stay at a much lean cost for consumer-facing products.<p>Here's how the whole stack looks today -<p>IDE - Google's AntiGravity (100% free + higher access if you use student ID)<br>
AI Documentation - SuperDocs (100% free & open source)<br>
Database - Supabase (Nano plan free, enough for basic needs)
Authentication - Stack Auth (Free upto 10K users)
LLM (AI Model) - OpenRouter or Gemini via AI Studio for testing and a custom tuned model by Unsloth AI for production. (You can fine-tune models using Unsloth literal in a Google Colab Notebook)
Version Maintenance/Distribution - Github/Gitlab (both totally free and open source)
Faster Deployment - Vercel (Free Tier Enough for Hobbyists)
Analytics - PostHog, Microsoft Clarity & Google Analytics (All 3 are free and independent for different tracking, I recommend using all of them)
That's the list devs! I know I might have missed something. If yes, just ask me up or list them up in the comments. If you have any questions related to something specific, ask me up as well.