Artist Portfolio Store: A Full-Stack Application Practice Integrating AI Content Generation

This article introduces a full-stack web application designed for independent artists, integrating portfolio display, e-commerce functions, and an AI content generation workflow based on LangGraph. It solves problems such as showcasing works, managing inventory, and generating social media copy. The tech stack includes Vue3, Node.js, Express, MongoDB, LangGraph, and Ollama, providing a practical solution for embedding AI in the creative industry.

Independent artists face challenges in effectively showcasing their works, managing inventory, and reaching buyers; traditional e-commerce platforms are too generic to reflect the uniqueness of art. The original intention of the project was to build an exclusive platform for independent artist brothers, considering the unique aspects of art sales (multiple derivative products) and the time-consuming problem of creating social media copy.

Core Function Modules: Public display site (two-layer structure: separate works and derivative products), admin workspace (session authentication, soft deletion, visibility toggle), catalog bundling API (batch management); AI Content Generation Workflow: Graph structure based on LangGraph, including input normalization, example loading, Ollama local model call, Zod validation retry; Tech Stack: Vue3+Vite frontend, Node.js+Express backend, MongoDB storage.

Data Model: Works, product types, products, product image entities, with relationships established via Mongoose; Technical Highlights: Password salted hashing (crypto module), session management (express-session), fault-tolerant retry for AI workflow; Deployment and Operation: Need to install Node.js, MongoDB, Ollama, start after configuring environment variables, supports development and production modes.

This project is a practical solution to the digital operation problems of independent artists, integrating traditional web functions with AI capabilities to lower the operational threshold. Its open-source nature allows community expansion, providing a reference for embedding AI in vertical domain applications.

Current Limitations: AI preferences are stored in memory (lost on restart), images only support URL references, single admin account, only Ollama local model; Improvement Directions: Persist preferences, add image upload, role permission system, support external AI APIs.