reddit · SideProject · 11-09 09:50 46 分 ← 返回列表

I built an open-source AI interview copilot that suggests answers while questions are being asked. Here's the demo

I've been building something called OpenKeet , and this is probably the feature I'm most excited to share. It's a desktop AI interview assistant, but I didn't want it to be just another app where you type interview questions into a chatbot. I wanted it to understand spoken questions and show useful suggestions while a conversation is happening. In the video, you can see OpenKeet: Listening to call audio. Recognizing questions as they're spoken. Displaying sugges

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：面试过程中难以实时组织高质量回答，而现有面试辅助工具多依赖付费云端服务，用户希望有本地、隐私安全、成本可控的实时提示方案

目标用户：频繁参加技术/行为面试的求职者（尤其软件工程师）、需要模拟面试的应届生、面试辅导机构

证据：One of my main goals is to avoid requiring a paid cloud infrastructure just to use the application, so I'm building around local AI options and user-configured providers. / Would you actually use something like this? / This demo uses fictional data and a local stand-in AI model. It's showing the desktop workflow, not claiming production-model accuracy or latency.

变现：开源核心 + 付费 Pro 桌面版（一次性买断或年订阅）：更好的本地模型预设、更低延迟、简历库与答案包管理；另可做面试题库/答案包的分发型付费市场抽成

MVP：Electron/Tauri 桌面壳 + 系统音频回环采集 + 本地 Whisper 流式转写 + 本地小模型（Ollama）生成要点式提示 + 透明置顶浮层，2 周可做出可演示版本；先只支持 macOS 或 Windows 单平台与一种面试语言

竞品：Final Round AI、Interview Coder、Cluely、Interviewer.AI、Ollama + Whisper 自建方案

摘要：作者开源了桌面 AI 面试副驾 OpenKeet，可实时监听通话音频、识别被问到的问题并在浮层给出建议答案，还包含简历准备、预生成答案、模拟面试和社区面试包。项目刻意围绕本地 AI 和用户自配 provider 构建，以避免付费云基础设施。帖子本身只是 demo 自荐，无收入或明确付费意愿数据，且该赛道已有多个成熟付费产品，竞争拥挤。