// the find
Snailclimb/interview-guide
基于 Spring Boot 4.1 + Java 21 + Spring AI 2.0 + PostgreSQL + pgvector + RustFS + Redis，实现简历智能分析、AI模拟面试、知识库RAG检索等核心功能。非常适合作为学习和简历项目，学习门槛低。
A Spring Boot 4.1 + Spring AI 2.0 demo project for AI-assisted job interview practice: resume parsing, mock interviews (text and voice), RAG knowledge bases, and calendar scheduling. It's explicitly positioned as a learning project and résumé line item for Chinese developers preparing for backend interviews.
Uses pgvector inside Postgres instead of a separate vector DB, which keeps the stack sane for a project at this scale. The voice interview pipeline is more complete than you'd expect — sentence-level concurrent TTS, server-side VAD, echo protection, and Micrometer instrumentation. Redis Stream for async resume analysis is a reasonable choice that avoids pulling in Kafka. The unified evaluation engine shared between text and voice interviews avoids duplicating the scoring logic.
JPA ddl-auto: update in development is a trap — schema drift between environments is how you lose data or ship silent bugs, and the README normalizes it. The voice interview latency problems (listed as known issues) are architectural: server-side audio relay and no WebRTC means this is demo quality, not something you'd put in front of real candidates. There is no auth layer described anywhere in the README or directory tree — the entire platform appears to be open-access, which matters if you deploy this anywhere public. The project is tightly coupled to Alibaba's DashScope API for both LLM and voice, so if you want to run it outside China or swap providers, the seams are thinner than the multi-provider settings page implies.