What is interview-guide?

InterviewGuide is an open source AI interview assistant platform built with Spring Boot 4, Java 21, Spring AI, PostgreSQL with pgvector, Redis, and a React and TypeScript frontend. It combines resume analysis, mock interviews in both text and voice, interview scheduling, and a knowledge base with retrieval augmented question answering into one system. The resume module accepts PDF, DOCX, DOC, and TXT files, processes them asynchronously through Redis Stream so a user can watch progress update in real time, and automatically retries failed analysis up to three times while detecting duplicate uploads by content hash. Results can be exported as a structured PDF report. The mock interview module uses skill files to drive question generation across more than ten tracks such as Java backend, frontend, Python, algorithms, system design, testing, and AI agents, each track defining its own scope, difficulty spread, and reference material. It avoids repeating questions a user has already answered in earlier sessions, allocates interview time across stages like self introduction and technical review, and supports configurable multi round follow up questions. A shared scoring engine evaluates both text and voice interviews the same way, producing batch scores, structured output, a summary, and a fallback path when scoring fails, so results stay comparable across sessions. Voice interviews run over WebSocket using a single API key for speech recognition, synthesis, and the language model, targeting a first response delay of around two hundred milliseconds through sentence level parallel speech synthesis and automatic pause detection. Known limits include noticeable end to end delay from server side audio conversion, occasional echo without headphones, and audio interruptions on weak networks. The knowledge base module supports uploading PDF, DOCX, and Markdown files, chunking and embedding them with pgvector, and answering questions through streaming responses. The system supports multiple model providers including DashScope, LM Studio, Kimi, DeepSeek, and GLM, switchable from a settings page without editing source code. All functionality is free and open source, with no separate paid tier. The full README is longer than what was shown.