# localdeploy.dev > Run your own AI assistant on hardware you already own localdeploy.dev helps home labs and small teams assess the hardware they already own, then deploy a local AI assistant with Cloudflare, self-hosting and containers. Free basics, fixed-scope deployment packages, clear boundaries. ## Pages - [The idle machine at home can run an AI assistant](/index.md): localdeploy.dev is one person's home-lab practice: local AI on hardware people already own, scoped deployment work, and notes from real builds. - [About localdeploy.dev](/about.md): An independent home-lab site: free guides, plus a small number of deployment projects. No affiliate links, no benchmark promises, no account handover. - [Cloudflare basics: domains, IPs, Pages, Workers and Tunnel](/cloudflare.md): One path diagram for Cloudflare: how domains and DNS work, where Pages and Workers fit, how Tunnel avoids inbound ports, and the order to debug a request. - [Contact: what to send, when to expect a reply](/contact.md): Send hardware, current system, goal and timeline to hello@localdeploy.dev. Also covers reply times, privacy and what this site does not do. - [Terms and privacy](/legal.md): Terms, liability limits, intellectual property, privacy and cookies for localdeploy.dev. No form collection, no online payment, no ads or tracking. - [Pricing: quoted by scope, not a fixed bundle](/pricing.md): Why local deployment is quoted by scope: hardware, current setup, migration and timing differ too much. Also covers the three modes, what changes a quote, payment and cancellation. No online payment yet. - [How a project runs](/process.md): From the first email to handover: scope, quote, build window, cutover, documentation and follow-up, plus a condensed sample log. - [Cancellation and refunds](/refund.md): Paid projects follow the written quote: what happens before work starts, after it starts and after delivery; which cases are excluded; how to ask. - [Home-lab security: do the few things that prevent most incidents](/security.md): The realistic risk for a home AI server is exposure, weak passwords, stale images and missing backups. Work through exposure, auth, secrets, containers, segmentation, backups and incident response. - [Deployment packages: six areas](/services.md): Traefik, Harbor, cloudflared, DeepSeek Harness, k3d/k3s and Grafana. Each area has its own page with scope, deliverables and exclusions; quotes by scope, no online payment yet. - [cloudflared tunnel: reach a service without opening a port](/services/cloudflared.md): Expose an internal service through Cloudflare Tunnel and Access when there is no public IP or port forwarding is unwanted. - [Grafana monitoring: let the problem ring first](/services/grafana.md): Prometheus, Loki and Grafana collection, alerting and retention; what gets delivered, and why to start with few accurate alerts. - [Harbor private registry: the same image on every machine](/services/harbor.md): When a self-hosted Harbor is worth it, what gets delivered, how storage and backups are planned, and when registry:2 is enough. - [DeepSeek Harness: running dsh in a container](/services/harness.md): dsh is the launcher for DeepSeek Harness. How its profiles and configuration are laid out, which volumes to mount, how upgrades and rollbacks work, and what is out of scope. - [k3d / k3s: from one Docker host to several nodes](/services/k8s.md): When lightweight Kubernetes is worth it, what gets delivered, when it is not, and how to pick between k3d and k3s. - [Traefik reverse proxy: one entrance for every service](/services/traefik.md): When you need Traefik, what gets delivered, what is out of scope, and how to move over from an existing Nginx setup. - [System and hardware choices: Debian, Proxmox, fnOS, TrueNAS or Flatcar?](/systems.md): There is no single best home AI server. Compare common systems by hardware support, GPU passthrough, storage, update model, learning curve and backups, then choose in a sensible order. - [Selected work](/work.md): A few local AI and self-hosting setups: dsh in containers, a Traefik cutover, a Harbor migration. Client and machine details are anonymised. - [家里闲置的机器,也能跑起 AI 助理](/zh.md): localdeploy.dev 是一个人的家庭实验室实践:把本地 AI 装到现有硬件上,接少量落地项目,记下实际排障的笔记。 - [关于 localdeploy.dev](/zh/about.md): 一个独立维护的家庭部署站点:写基础,也接少量落地项目。没有联盟返佣,不承诺跑分,不接管账号。 - [Cloudflare 基础:域名、IP、Pages、Workers 与 Tunnel](/zh/cloudflare.md): 用一张路径图理解 Cloudflare:域名和 DNS 怎么工作,Pages 和 Workers 分别适合什么,Tunnel 如何避免公网入站,以及出问题时的排查顺序。 - [联系:发什么、多久回复](/zh/contact.md): 把硬件配置、现有系统、目标和时间要求发到 hello@localdeploy.dev。也说明回复节奏、隐私边界,以及哪些事不做。 - [条款与隐私](/zh/legal.md): localdeploy.dev 的服务条款、责任边界、知识产权、隐私和 Cookie 说明。不收集表单内容,不在线收款,不接广告和统计追踪。 - [定价:按范围报,不卖固定套餐](/zh/pricing.md): 为什么本地部署按范围报价:硬件、现状、迁移量和时间窗口差得太多。也说明三种做法、报价看什么、怎么付款和取消;现在不开在线支付。 - [项目是怎么跑的](/zh/process.md): 从第一封邮件到交付:范围、报价、实施窗口、切换、文档和答疑,附一份脱敏的项目记录。 - [退款与取消](/zh/refund.md): 付费项目按报价单取消和退款:没开始、开始了没交付、已经交付分别怎么处理,哪些情况不适用,以及怎么申请。 - [家庭实验室的网络安全:先做能挡住大多数事故的事](/zh/security.md): 家庭 AI 服务器最现实的风险不是高级攻击,而是暴露面、弱口令、过期镜像和缺失备份。按暴露面、认证、密钥、容器、网络分割、备份和事件响应逐项检查。 - [付费项目:六个方向](/zh/services.md): Traefik、Harbor、cloudflared、DeepSeek Harness、k3d/k3s、Grafana。每个方向单独一页写范围、交付和不做什么;报价按范围来,现在不开在线支付。 - [cloudflared 内网穿透:不开端口,把服务接出去](/zh/services/cloudflared.md): 没有公网 IP 或不想做端口映射时,用 Cloudflare Tunnel 和 Access 暴露内网服务的交付范围与排查顺序。 - [Grafana 监控:让问题先响一声](/zh/services/grafana.md): Prometheus、Loki 和 Grafana 的采集、告警和保留策略;交付什么,以及为什么先做少而准的告警。 - [Harbor 私有镜像仓库:多台机器拉同一个版本](/zh/services/harbor.md): 自建 Harbor 的适用场景、交付内容、存储与备份规划,以及什么时候 registry:2 就够用。 - [DeepSeek Harness:把 dsh 装进容器长期跑](/zh/services/harness.md): dsh 是 DeepSeek Harness 的启动器。这一页说明它怎么组织配置、容器化时要挂哪些卷、升级和回滚怎么做,以及哪些不在交付范围内。 - [k3d / k3s:从一台 Docker 主机走到多节点](/zh/services/k8s.md): 什么时候值得上轻量 Kubernetes、交付哪些东西、什么时候不该上,以及 k3d 和 k3s 怎么选。 - [Traefik 反向代理:一个入口管住所有服务](/zh/services/traefik.md): 什么时候需要 Traefik、交付哪些配置、不包含什么,以及从现有 Nginx 迁过去的顺序。 - [系统与硬件选型:Debian、Proxmox、fnOS、TrueNAS 还是 Flatcar?](/zh/systems.md): 家庭 AI 服务器没有唯一正确答案。按硬件兼容性、GPU 直通、存储、更新策略、学习曲线和备份方式比较常见系统,给出可执行的选型顺序。 - [做过的项目](/zh/work.md): 三个本地 AI 和自托管的项目:dsh 容器化、Traefik 迁移、Harbor 镜像仓库。客户和机器细节都隐去了。 ## Delivery and payment - No online checkout is live at this stage. Paid deployment packages are arranged by email. - Scope, timing and payment terms are confirmed in writing before work starts (hello@localdeploy.dev). - The site never handles card data and does not create an account or subscription. - Refund terms: [Cancellation and refunds](/refund.md) - Refund terms: [退款与取消](/zh/refund.md) ## Contact - hello@localdeploy.dev ## Optional - [Full site text](/llms-full.txt): every public page and product body in one file. - [Agent capabilities](/skill.md): capabilities, delivery flow and boundaries. - [Discovery metadata](/.well-known/ard.json): ARD entry for this store. - Markdown negotiation: send `Accept: text/markdown` to any page URL to receive its Markdown mirror.