AI

AI Comment Moderation: Zero-Cost Automatic Content Filtering with Gemini 2.0 Flash

A blog system integrates Google Gemini 2.0 Flash (permanently free) to achieve AI-powered automatic comment moderation. Covers a three-tier review mechanism, asynchronous BullMQ queue processing, automatic reply prompt design, and non-blocking architecture design points.

Next.js Blog Bilingual System: Zero-Intrusion Automatic Translation Practice Based on Gemini AI

AI Translation Task Monitoring Dashboard: 1309 Lines Complete Engineering Implementation

Build an enterprise-level AI translation monitoring center—from BullMQ queue real-time monitoring, multi-dimensional progress tracking, to task history management and error alerts

AI Translation Issue Detection Panel: Engineering Practices for Batch Repairing Multilingual Content

How to build an 822-line translation quality detection panel—automatically scanning translation issues in articles/categories/tags, supporting batch repair and language filtering

Gemini AI Production-Grade Integration: Four-Layer Circuit Breaker Degradation and Triple Quota Control

An in-depth analysis of the production-grade encapsulation of @google-cloud/vertexai (Gemini 2.5 Flash) in a NestJS project, covering core implementations such as a four-layer service level degradation system, triple quota control, circuit breaker pattern, and exponential backoff retry.

Translation System Evolution Chronicle: From DeepSeek Migration to 429 Rate Limiting, Speed Optimization, and Detection Enhancement

This article documents the ongoing operational challenges of the blog translation system after the Provider migration, covering DeepSeek speed optimization, 429 rate limiting handling, category/tag translation detection fixes, and the complete implementation of admin backend translation enhancement features.

LanguageDetectionService: Hybrid Strategy Language Detection Engine

The JoyMini API's language detection service employs a three-layer hybrid strategy using the franc-min library, heuristic rules, and AI fallback, supporting 191 language recognition for blog multilingual translation and automatic matching of user content language.

AI Multi-Provider Abstraction Layer: Unified Access and Automatic Failover for Gemini + Groq + DeepSeek

The JoyMini API's AI service has evolved from a single Gemini Provider to a multi-Provider abstraction architecture, supporting unified access, automatic failover, Provider-level key rotation, and Admin UI visual configuration for three AI providers: Gemini, Groq, and DeepSeek.