Miantuan AI

📋 Project Overview

Miantuan is a job-seeker product that combines practice, live assistance, prep, and monetization in one place: AI mock interviews, AI assistance during real interviews, a prep assistant, resume + knowledge base (“cheat sheet”), plus credits, checkout, campaigns, and redemptions.

Live site: https://miantuan.knowfuture.com.cn/

🚀 Key Features

Product

• AI mock interviews: Start from a preset role or custom JD only; multi-turn room sessions with scoring and review and rescore; an in-room assistance panel shares the same quick-answer stack as live assistance. List/detail/cancel/delete for records; prompts if resume is missing.
• AI-assisted live interviews: For real online interviews. Configure language and stack; single PC or PC + phone dual-device (secondary device joins by scan); screen sharing; optional screenshot solving with a separate upload path for large captures; streamed quick answers (brief/detail, intent combined flows, web-augmented answers, etc.).
• Prep assistant: Upload resume and target company/role to generate Q&A (resume-personalized and web-sourced questions); draft box; streamed AI answer refinement; save/bookmark Q&A; company/BU/role data for filtering.
• Cheat sheet (knowledge base): Uploads, chunking, and full-text search.
• Resume: Upload and parsing for mock and prep flows.
• Account & growth: SMS verification, WeChat login; free credits, campaigns/popups, redeem codes; orders & products; WeChat Pay and Alipay.
• Console: Sidebar to mock interviews, assistance, prep assistant, resume, cheat sheet, free credits; footer balance & top-up; in-progress interviews can be rejoined or closed.
• Admin: Allow-listed operators maintain the mock interview question bank and related ops content.
• Written exam: Backend supports multimodal parsing of written-exam screenshots; desktop client update copy points users to written-exam features; the web “written exam mode” sidebar is coming soon—see the live site.

Engineering (summary)

Spring Boot WebFlux on the server; Volcengine Doubao as the main LLM with streaming, WebSocket rooms, and multimodal/vision paths; LangChain4j for parts of orchestration; analytics for iteration. Angular 19 on the client with SSE, Microsoft Speech SDK, and Markdown editing/viewing.

💻 Project Detail (user journey)

After sign-in, users can practice in mock interviews and read scores; run live assistance with sharing, dual-device, screenshots, and quick answers; use the prep assistant to generate and polish Q&A; maintain cheat-sheet material; and use credits, top-up, and campaigns. Operators curate content in the admin area.

📊 Project Impact

A public, end-to-end product spanning mock practice, live assistance, prep, and payments/login, with a stack that covers dialogue & multimodal models, real-time comms, search, and object storage.

🛠️ Technology Stack

Frontend:
  - Angular 19
  - TypeScript
  - NG-Zorro / Angular Material
  - Microsoft Cognitive Services Speech SDK
  - SSE (Fetch Event Source)
  - Vditor / ngx-markdown / Prism

Backend:
  - Java 21
  - Spring Boot 3 + Spring WebFlux
  - MyBatis-Plus
  - MySQL + Flyway
  - Redis
  - MongoDB Reactive
  - Volcengine Ark / Doubao (LLM, streaming, multimodal)
  - LangChain4j
  - Elasticsearch
  - Aliyun OSS
  - Apache POI
  - WebSocket (interview rooms)
  - Java-WebSocket (speech-related integration)
  - JWT

Payments & Auth:
  - WeChat Pay / Alipay
  - Aliyun SMS
  - WeChat login

Product & Ops:
  - Multi-environment frontend builds (local/dev/prod)
  - Activity configs, redeem codes, orders & products
  - Admin question-bank tooling

Feature availability and “coming soon” items follow the live product.