📡 Idea Radar 每天 8:00 / 18:00 推送 · 18:00 生成精选
reddit · SideProject · 51 分 ← 返回列表

I built an encrypted relay so I can use my home LocalLLM from my work laptop without leaking code.

Public cloud AI is often banned on confidential client work, and corporate firewalls block custom ports and SSH tunnels. I built k-lani-link to solve this. It creates an end-to-end encrypted bridge between your work laptop and your private AI machine (home GPU or lab server or whatever). It connects directly via authenticated WebRTC (P2P), with an ordinary HTTPS relay fallback (Cloudflare Worker, PHP/WordPress, or Rust) for strictly firewalled networks where direct UDP is bl

五维评分

付费意愿
45
需求真实度
55
个人可落地
55
竞争空间
35
新鲜度
70

AI 分析

痛点:处理保密客户项目的开发者无法使用公有云AI,公司防火墙又封锁自定义端口和SSH隧道,因此很难安全地从工作电脑访问家中或实验室的本地大模型。

目标用户:处理保密客户项目、内网管控严格的开发者、顾问、法务和医疗等数据敏感行业的小团队

证据:Public cloud AI is often banned on confidential client work, and corporate firewalls block custom ports and SSH tunnels. / Pricing: Free for production under $100k annual revenue, or €4.99 one-time.

变现:开源核心,并按席位向企业版收取订阅费(团队SSO、审计日志、中继托管、策略管控);个人版维持一次性低价授权作为获客入口。

MVP:用两周做一个单二进制CLI加托管中继:一端在本机启动OpenAI兼容代理,通过WebRTC打洞直连家庭GPU,UDP被封时自动回退到HTTPS中继;后端只对接Ollama或llama.cpp,一条命令启动,不保存提示日志。

竞品:Tailscale、Cloudflare Tunnel、ngrok、Open WebUI 远程访问、Ollama 自带暴露端口

摘要:作者用Rust写了k-lani-link,用认证后的WebRTC P2P加HTTPS中继(Cloudflare Worker/PHP/Rust回退)打通工作笔记本与家里本地大模型,兼容任意OpenAI客户端与Ollama/vLLM/llama.cpp等引擎,号称无遥测、无提示日志。定价为年收入10万美元以下免费、否则€4.99一次性买断,价格极低且无评论或收入数据支撑,付费意愿尚未验证;网络层需求已大量被Tailscale/Cloudflare Tunnel覆盖,差异化空间有限。