开源的AI面试官-Liftoff Interviews

Liftoff Interviews是一款面试准备的开源工具，可为模拟面试提供 AI 反馈。可让 AI 进行行为和技术面试，并给出能力评估，帮助用户改善不足之处，提高面试通过率。用户可根据自身能力来选择面试题难度。基于TypeScript编写，遵守MIT开源协议。

Liftoff 使用 FFmpeg 将原始视频转码为 MP3。然后，直接将音频发送到 OpenAI 的 Whisper 端点进行转录，使用 OpenAI 的 gpt-3.5-turbo 从边缘流式传输反馈。

源代码：https://github.com/Tameyer41/liftoff

---------

Mock Interview Simulator with AI-Powered Feedback.

demo.useliftoff.com

Liftoff Interviews

Mock Interview Simulator with AI-Powered Feedback

Introduction · One-click Deploy · Tech Stack + Features · Author

Introduction

Liftoff is an interview preparation tool that provides AI feedback on your mock interviews.

One-click Deploy

You can deploy this template to Vercel with the button below:

You can also clone & create this repo locally with the following command:

npx create-next-app liftoff --example "https://github.com/Tameyer41/liftoff"

Tech Stack + Features

Frameworks

• Next.js – React framework for building performant apps with the best developer experience

Platforms

• Vercel – Easily preview & deploy changes with git
• Upstash - Serverless Data Platform (here using serverless Redis for rate limiting)

UI

• Tailwind CSS – Utility-first CSS framework for rapid UI development
• Framer Motion – Motion library for React to animate components with ease
• ImageResponse – Generate dynamic Open Graph images at the edge
• HeadlessUI - Completely unstyled, fully accessible UI components, designed to integrate beautifully with Tailwind CSS

Code Quality

• TypeScript – Static type checker for end-to-end typesafety
• Prettier – Opinionated code formatter for consistent code style
• ESLint – Pluggable linter for Next.js and TypeScript

Miscellaneous

• FFMPEG.WASM – Transcode video/audio files
• React Webcam - Webcam component for React
• Stripe Gradient Animation - @jordienr released a Mesh Gradient that uses WebGL and animates a beautiful gradient

How it all works

Liftoff uses FFmpeg to transcode the raw video into MP3. Chrome, Safari, and Firefox all record with different codecs, and FFmpeg is great for standardizing them.

We then send the audio directly to be transcribed by OpenAI's Whisper endpoint, and then stream feedback from the edge using OpenAI's gpt-3.5-turbo.

from https://github.com/Tameyer41/liftoff