Show HN: 用 AI 边缘代理替代 Wix，从此不再需要雇佣初级员工
8 分•作者: axotopia•3 天前
我经营一家建筑设计咨询公司。我厌倦了每月向 Wix 支付 40 美元，却只能得到一个无法回答简单服务问题的宣传册，而且我还要花几个小时处理相同的常见问题解答。<p>所以我把它全部砍掉，花了 4 个月时间构建了一个“对话机器人”：<a href="https://axoworks.com" rel="nofollow">https://axoworks.com</a><p>技术栈完全是拼凑起来的：Netlify 的 10 秒无服务器超时迫使我将代理分成三部分：大脑（边缘服务器）、双手（浏览器）和声音（边缘服务器）。我已经 30 年没有写过代码了。这算得上是前进 3 步，后退 2 步，很大程度上是 AI 引导的。<p>证明它有效的战斗：两周前，一位持证建筑师攻击了这个机器人，试图证明我的商业模式会损害这个行业。AI（DeepSeek-R3）完全驳斥了他的论点。过程非常辛辣，令人捧腹。<p>日志：<a href="https://logs.axoworks.com/chat-architect-vs-concierge-v147.html" rel="nofollow">https://logs.axoworks.com/chat-architect-vs-concierge-v147.h...</a><p>一些惨痛的教训：<p>* Web Speech API 运行良好，直到有人在没有切换语言模式的情况下说中文。然后它会强行吐出英语的语音乱码。这仍然是个难题。<p>* 责任是致命的。如果 AI 虚构了建筑规范条款？我们就完蛋了。保险公司不会碰我们。<p>* 我们发布审计日志以保持诚实，并确保系统保持稳固。<p>审计：<a href="https://logs.axoworks.com/audit-2026-02-19-v148.html" rel="nofollow">https://logs.axoworks.com/audit-2026-02-19-v148.html</a><p>最难的部分是正确把握意图：让一个 LLM 在与房主交流时无缝切换到温暖的负责人语气，而在被同行攻击时切换到防御性的斗牛犬模式。这花了 2.5 个月的调整时间。<p>我们通过一个“Eager RAG”技巧（预先获取猜测）来消耗大量 token，只是为了提高响应速度。我还删除了“必要的”持久数据库——只有不到 5% 的访问者会再次访问，所以何必呢？如果客户在查询过程中中断，他们的会话就会消失。没有服务器端队列。<p>重点：让我能够与经验丰富的专业人士网络合作，并精简冗余环节。<p>试试看，能否把它搞坏。我会在评论区等着。继续
查看原文
I run a building design consultancy. I got tired of paying Wix $40/month for a brochure that couldn’t answer simple service questions, and me wasting hours on the same FAQs.<p>So I killed it all and spent 4 months building a 'talker': <a href="https://axoworks.com" rel="nofollow">https://axoworks.com</a><p>The stack is completely duct-taped: Netlify’s 10s serverless timeout forced me to split the agent into three pieces: Brain (Edge), Hands (Browser), and Voice (Edge). I haven’t coded in 30 years. This was 3 steps forward, 2 steps back, heavily guided by AI.<p>The fight that proved it worked: 2 weeks ago, a licensed architect attacked the bot, trying to prove my business model harms the profession. The AI (DeepSeek-R3) completely dismantled his arguments. It was hilariously caustic.<p>Log: <a href="https://logs.axoworks.com/chat-architect-vs-concierge-v147.html" rel="nofollow">https://logs.axoworks.com/chat-architect-vs-concierge-v147.h...</a><p>A few battle scars:<p>* Web Speech API works fine, right up until someone speaks Chinese without toggling the language mode. Then it forcefully spits out English phonetic gibberish. Still a headache.<p>* Liability is the killer. Hallucinate a building code clause? We’re dead. Insurance won’t touch us.<p>* We publish the audit logs to keep ourselves honest and make sure the system stays hardened.<p>Audit: <a href="https://logs.axoworks.com/audit-2026-02-19-v148.html" rel="nofollow">https://logs.axoworks.com/audit-2026-02-19-v148.html</a><p>The hardest part was getting the intent right: making one LLM pivot seamlessly from a warm principal’s tone with a homeowner, to a defensive bulldog when attacked by a peer. That took 2.5 months of tuning.<p>We burn through tokens with an 'Eager RAG' hack (pre-fetching guesses) just to improve responsiveness. I also ripped out the “essential” persistent DBs—less than 5% of visitors ever return, so why bother? If a client drops mid-query, their session vanishes. No server-side queues.<p>The point: To let me operate with a network of seasoned pros, and trim the fat.<p>Try to break it. I’ll be in the comments. Kee