SuperAPI AI Directory

cheetah

Cheetah is an AI-powered macOS app designed to assist users with software engineering interview practice. It provides real-time speech transcription and real-time coding platform integration, helping users get hints and answers during mock interviews. Its core dependencies are Whisper for real-time audio transcription (running locally on whisper.cpp), and GPT-4 for generating hints and answers. Users need to bring their own OpenAI API key, and a relatively new Mac with an Apple chip is required for best performance. The app includes three buttons: Answer, Refine, and Analyze, and supports a Firefox browser extension to analyze code and logs in coding environments. Note: This app is for mock interviews only, may generate incorrect or inappropriate answers, and users bear full responsibility for the information provided by the app.

Open website

About this tool

Cheetah is an AI-based macOS app designed specifically for software engineering interview practice, providing real-time coaching and real-time coding platform integration.

Main capabilities: The app uses Whisper for real-time audio transcription and GPT-4 for generating hints and answers. Whisper runs locally on the system via Georgi Gerganov's whisper.cpp. Users need to bring their own OpenAI API key. For best performance, a relatively new Apple chip Mac is required.

Usage: Requires macOS 13.1 or later. Before building Cheetah, whisper.cpp needs to be checked out to ../whisper.cpp, and the SDL2 library needs to be installed. Regarding audio driver setup: for best results, ensure the audio input can capture the voices of both parties in the conversation. When using video chat apps such as Zoom or Google Meet, this can be achieved through the free audio loopback driver BlackHole, following the multi-output device setup instructions, and be careful not to use the loopback device as the input for the video chat app. Open the app and select the audio input to begin real-time transcription; the transcribed content snippets will be displayed below the audio input selector. Note: Running the app in debug mode will cause very slow audio transcription performance.

The interface provides three buttons: Answer generates an answer to the interviewer's question; Refine updates an existing answer, which is useful when the interviewer provides additional constraints or clarifications; Analyze analyzes code and logs from the real-time coding environment in the browser, and requires the browser extension. In addition, users can also select (highlight) part of a generated answer and then click Refine to get more details.

Browser extension installation: Currently only Firefox is supported. The steps include: adding the domain of the real-time coding platform to the matches in ./extension/manifest.json; visiting about:debugging; clicking "This Firefox"; clicking "Load Temporary Add-on"; selecting ./extension/manifest.json.

Use cases and limitations: Cheetah is for mock interviews only, may generate incorrect or inappropriate answers, and users bear full responsibility for the information provided by the app.