An AI interviewer built from Redis streams and good intentions

What it does

InterviewGuide is a full-stack AI interview platform. The Java 21 backend uses Spring Boot 4.0.1, Spring AI 2.0.0-M4, and virtual threads to parse resumes, conduct mock interviews, and serve a RAG knowledge base over PostgreSQL and pgvector. A React frontend provides calendars, chat, and a control panel for swapping LLM providers without touching code.

The interesting bit

The author deliberately keeps the stack minimal: Redis Stream handles async resume analysis and vectorization instead of Kafka, and PostgreSQL with pgvector doubles as both relational store and vector database. The voice pipeline is surprisingly deep—server-side VAD, sentence-level concurrent TTS, and echo protection—but the README openly notes it still struggles with latency and audio leakage without headphones.

Key highlights

• Unified evaluation engine for text and voice interviews, with structured scoring and PDF report export via iText 8.
• Skill-driven question generation using SKILL.md files for 10+ interview tracks, plus automatic deduplication against past sessions.
• Interview scheduling that parses Feishu, Tencent Meeting, and Zoom invites with a rule-plus-AI engine, then renders them in a drag-and-drop calendar.
• Runtime model switching and encrypted API-key storage in ~/.interview-guide/, supporting DashScope, DeepSeek, Kimi, GLM, and LM Studio.
• Async processing with Redis Stream, including automatic retries and content-hash deduplication for resume analysis.

Caveats

• The voice interview module currently has noticeable end-to-end delay, single-voice TTS, and echo leakage without headphones; WebRTC and client-side VAD are listed as future work.
• Several major integrations—such as connecting mock interviews directly to the knowledge base—remain on the TODO list.

Verdict

Great for Java developers who want a credible, resume-worthy project that demonstrates Spring AI, RAG, and async architecture without hiding behind buzzwords. Skip it if you need a polished, production-ready voice agent today.