ragent

Ragent is a Java-based, enterprise-grade Agentic RAG platform built with Spring AI that covers the full pipeline from document ingestion to intelligent question answering. It implements parallel multi-channel retrieval with deduplication and reranking, a tree-structured intent classifier that requests clarification when confidence falls below a threshold, query rewriting for multi-turn conversations, session memory management, and MCP tool integration that automatically invokes business tools for non-knowledge intents. The platform includes a model scheduling layer with health checks, first-packet latency probing, and automatic failover so that individual model outages do not interrupt service. It connects to LLMs through a custom scheduling engine rather than LangChain4j or Spring AI abstractions, a deliberate design choice the project documents in detail. The primary audience is Java backend developers transitioning into AI engineering roles, and the repository explicitly frames the codebase as resume-ready material for candidates who need to demonstrate production-grade RAG and Agent experience in technical interviews.

Lo que la gente pregunta sobre ragent