reddit · SideProject · 10-18 11:50 50 分 ← 返回列表

I built a live status page that tracks AI model outages

After getting blocked by Gemini API 503s during a client demo (and having no idea if it was me or them), I built ModelDown — https://modeldown.si It checks OpenAI, Anthropic, Gemini, xAI, DeepSeek, Groq, Mistral, Cohere, Together, and OpenRouter every 5 minutes via their status APIs and live probes. There's also a free weekly outage roundup email. Would love some feedback, especially from anyone who's been burned by a model outage. What would make this actually useful to you? subm

五维评分

付费意愿

需求真实度

个人可落地

竞争空间

新鲜度

AI 分析

痛点：调用多家大模型 API 时，遇到 503/限流/宕机无法快速判断是自己代码问题还是服务商故障，导致演示或线上业务翻车

目标用户：依赖多个 LLM API 的独立开发者、AI 应用小团队、做 agent/RAG 产品的工程师与 SRE

证据：After getting blocked by Gemini API 503s during a client demo (and having no idea if it was me or them), I built ModelDown... Would love some feedback, especially from anyone who's been burned by a model outage. What would make this actually useful to you?

变现：免费公开状态页引流 + 付费订阅（$9–29/月）：Slack/Discord/Webhook 实时告警、多供应商宕机历史与 SLA 报表、自动 fallback 路由代理（OpenAI 挂了自动切 Anthropic）、团队席位

MVP：2 周内做：聚合 10 家官方 status API + 每 5 分钟轻量探测（发一条最小 token 请求测真实可用性，而非只看官方状态），存时序数据，提供公开状态页、单个 webhook 告警和可嵌入的 status badge；第二版再加 OpenAI 兼容的 fallback 网关按可用性自动切模型

竞品：StatusGator、各厂商官方 status page（OpenAI/Anthropic 等）、OpenRouter 自带可用性面板、Downdetector 类通用监控站

摘要：作者在客户演示时被 Gemini API 503 打断，分不清是自己还是服务商的问题，于是做了 ModelDown，每 5 分钟通过官方 status API 加实测探针监控 OpenAI、Anthropic、Gemini 等 10 家模型服务，并提供免费每周宕机汇总邮件。帖子只是求反馈，没有任何付费意愿或收入数据，但“多模型可用性监控 + 自动切换”对依赖 LLM 的小团队是真实痛点，适合单人低成本切入。