JeecgBoot

Converts single-sentence natural language descriptions into complete working systems by leveraging LLM integration (via Spring-AI and LangChain4j) to interpret intent, generate data models, and orchestrate the OnlineCoding visual configuration engine. The system uses prompt engineering to extract entity definitions, relationships, and business rules from unstructured text, then maps these to the @jeecg/online form designer and database schema generator, producing executable applications without manual coding.

Provides a unified abstraction layer (via Spring-AI and jeecg-boot-module-airag) for managing multiple LLM providers (OpenAI, Deepseek, Anthropic, local Ollama instances) with dynamic model selection, fallback routing, and provider-agnostic prompt execution. The system maintains a model registry in the database, supports hot-swapping between providers without code changes, and includes cost tracking and usage analytics per model.

Implements a fine-grained authorization system combining role-based access control (RBAC) for feature/API access with row-level security (RLS) for data filtering. The system stores roles, permissions, and data permission rules in the database, evaluates permissions at the API layer using Spring Security interceptors, and applies row-level filters at the SQL query level using MyBatis-Plus interceptors. Data permissions can be based on user attributes (department, region) or custom business rules.

Supports microservices deployment using Spring Cloud Alibaba 2023.0.3.3 with Nacos for service discovery, configuration management, and load balancing. The system provides API Gateway routing, circuit breaker patterns via Sentinel, distributed tracing via Skywalking, and inter-service communication via Feign clients. Services can be deployed independently and registered with Nacos for dynamic discovery.

Implements distributed transaction support using Seata (Alibaba's distributed transaction framework) with AT (Automatic Transaction) mode for transparent transaction coordination across multiple databases. The system maintains transaction logs, supports rollback on failure, and ensures eventual consistency across services. Seata integrates with Spring Transaction management for seamless distributed transaction handling.

Packages the Vue3 frontend as an Electron desktop application with offline capabilities via PWA (Progressive Web App) service workers. The system caches critical assets and API responses, syncs data when connectivity is restored, and provides native desktop features (file system access, system tray integration). The Electron wrapper communicates with the Spring Boot backend via HTTP/WebSocket, supporting both online and offline modes.

Automatically generates OpenAPI 3.0 specifications from Spring Boot controller annotations using Springdoc-OpenAPI, exposing interactive Swagger UI for API exploration and testing. The system introspects REST endpoints, request/response schemas, and validation rules, generating comprehensive API documentation without manual specification writing. Documentation is updated automatically when code changes.

Implements a complete Retrieval-Augmented Generation pipeline (jeecg-boot-module-airag) that ingests documents (PDF, Word, text), chunks them using configurable strategies, generates embeddings via LLM providers, stores vectors in a vector database, and retrieves relevant context for LLM queries using semantic similarity search. The system uses LangChain4j for orchestration, supports multiple embedding models, and includes document metadata indexing for hybrid search (semantic + keyword filtering).

Provides a visual workflow designer (AIFlow module) that allows users to compose AI-driven processes by connecting nodes (LLM calls, data transformations, conditional logic, API calls) in a DAG structure. The system executes workflows using an event-driven engine that processes nodes asynchronously, maintains execution state, supports branching/looping, and integrates with the knowledge base and model management systems. Workflows are stored as JSON configurations and can be triggered via REST API or scheduled execution.

Implements a chat service (jeecg-boot-module-airag) that maintains conversation history, manages token-aware context windows, integrates with the knowledge base for RAG-augmented responses, and supports skill/tool calling for external API integration. The system uses LangChain4j for conversation memory management, supports multiple conversation threads per user, and includes message persistence for audit and analytics.

Generates complete full-stack code (Vue3 frontend + Spring Boot backend + MyBatis-Plus entities) from database schemas and form definitions using the CodeGenerateUtil class and OnlineCoding configuration. The generator introspects database tables, infers CRUD operations, generates typed entity classes, REST endpoints with pagination/filtering, and Vue3 form components with validation rules. Generated code is immediately executable and can be customized via code templates.

Provides a drag-and-drop form designer (@jeecg/online package) that allows users to compose forms by selecting field types, configuring validation rules, and defining conditional visibility/enable logic without coding. The designer generates form JSON configurations that are rendered at runtime by Vue3 components, supporting complex field types (date pickers, selects with dynamic options, file uploads, rich text editors) and cross-field validation rules.

Integrates with Activiti and Flowable workflow engines to provide a visual BPMN 2.0 process designer that allows users to model business processes (tasks, gateways, events, subprocesses) and execute them with human task assignment, process variables, and audit trails. The system stores process definitions in the database, manages task queues, supports process versioning, and integrates with the user/role system for task assignment and approval workflows.

Supports multiple database backends (MySQL, PostgreSQL, Oracle, SQL Server, H2) with automatic SQL dialect translation via MyBatis-Plus and database-specific configuration. The system uses ShardingSphere for transparent data sharding across multiple database instances, supporting both range-based and hash-based sharding strategies. Database migrations are managed via Flyway with dialect-specific migration scripts.

Provides a plugin architecture (jeecg-boot-plugin module) that allows developers to extend platform functionality by implementing standard plugin interfaces and registering them via Spring component scanning. The system supports Model Context Protocol (MCP) for standardized tool/skill integration, enabling plugins to expose capabilities as callable functions with schema-based validation. Plugins can hook into lifecycle events (startup, shutdown, data changes) and access core services via dependency injection.