Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Higress-RAG: A Holistic Optimization Framework for Enterprise Retrieval-Augmented Generation via Dual Hybrid Retrieval, Adaptive Routing, and CRAG

Weixi Lin

TL;DR

The results demonstrate that by optimizing the entire retrieval lifecycle - from pre-retrieval query rewriting to post-retrieval corrective evaluation - the Higress RAG system offers a scalable, hallucination-resistant solution for enterprise AI deployment.

Abstract

The integration of Large Language Models (LLMs) into enterprise knowledge management systems has been catalyzed by the Retrieval-Augmented Generation (RAG) paradigm, which augments parametric memory with non-parametric external data. However, the transition from proof-of-concept to production-grade RAG systems is hindered by three persistent challenges: low retrieval precision for complex queries, high rates of hallucination in the generation phase, and unacceptable latency for real-time applications. This paper presents a comprehensive analysis of the Higress RAG MCP Server, a novel, enterprise-centric architecture designed to resolve these bottlenecks through a "Full-Link Optimization" strategy. Built upon the Model Context Protocol (MCP), the system introduces a layered architecture that orchestrates a sophisticated pipeline of Adaptive Routing, Semantic Caching, Hybrid Retrieval, and Corrective RAG (CRAG). We detail the technical implementation of key innovations, including the Higress-Native Splitter for structure-aware data ingestion, the application of Reciprocal Rank Fusion (RRF) for merging dense and sparse retrieval signals, and a 50ms-latency Semantic Caching mechanism with dynamic thresholding. Experimental evaluations on domain-specific Higress technical documentation and blogs verify the system's architectural robustness. The results demonstrate that by optimizing the entire retrieval lifecycle - from pre-retrieval query rewriting to post-retrieval corrective evaluation - the Higress RAG system offers a scalable, hallucination-resistant solution for enterprise AI deployment.

Higress-RAG: A Holistic Optimization Framework for Enterprise Retrieval-Augmented Generation via Dual Hybrid Retrieval, Adaptive Routing, and CRAG

TL;DR

The results demonstrate that by optimizing the entire retrieval lifecycle - from pre-retrieval query rewriting to post-retrieval corrective evaluation - the Higress RAG system offers a scalable, hallucination-resistant solution for enterprise AI deployment.

Abstract

The integration of Large Language Models (LLMs) into enterprise knowledge management systems has been catalyzed by the Retrieval-Augmented Generation (RAG) paradigm, which augments parametric memory with non-parametric external data. However, the transition from proof-of-concept to production-grade RAG systems is hindered by three persistent challenges: low retrieval precision for complex queries, high rates of hallucination in the generation phase, and unacceptable latency for real-time applications. This paper presents a comprehensive analysis of the Higress RAG MCP Server, a novel, enterprise-centric architecture designed to resolve these bottlenecks through a "Full-Link Optimization" strategy. Built upon the Model Context Protocol (MCP), the system introduces a layered architecture that orchestrates a sophisticated pipeline of Adaptive Routing, Semantic Caching, Hybrid Retrieval, and Corrective RAG (CRAG). We detail the technical implementation of key innovations, including the Higress-Native Splitter for structure-aware data ingestion, the application of Reciprocal Rank Fusion (RRF) for merging dense and sparse retrieval signals, and a 50ms-latency Semantic Caching mechanism with dynamic thresholding. Experimental evaluations on domain-specific Higress technical documentation and blogs verify the system's architectural robustness. The results demonstrate that by optimizing the entire retrieval lifecycle - from pre-retrieval query rewriting to post-retrieval corrective evaluation - the Higress RAG system offers a scalable, hallucination-resistant solution for enterprise AI deployment.
Paper Structure (35 sections, 2 equations, 5 figures, 2 tables)

This paper contains 35 sections, 2 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. The Imperative of Retrieval-Augmented Generation
  3. Limitations of Standard RAG Architectures
  4. The Higress Solution: Full-Link Optimization
  5. Related Work and Theoretical Foundations
  6. Evolution of RAG Systems
  7. The Model Context Protocol (MCP)
  8. Corrective Retrieval Augmented Generation (CRAG)
  9. Hypothetical Document Embeddings (HyDE)
  10. System Architecture
  11. Layer 1: The MCP Server Layer
  12. Layer 2: The RAG Client Layer
  13. Layer 3: Enhanced RAG Components Layer
  14. Layer 4: Infrastructure Layer
  15. Key Technologies and Methodologies
  16. ...and 20 more sections

Figures (5)

  • Figure 1: The Layered Architecture of Higress-RAG MCP Server. It illustrates the separation of concerns between the MCP Server Layer, RAG Client Layer, and Enhanced Components.
  • Figure 2: The Holistic Optimization Pipeline. The process flows dynamically from Semantic Caching to Adaptive Routing, followed by Hybrid Retrieval and Corrective Evaluation.
  • Figure 3: The Corrective RAG (CRAG) Decision Logic. The evaluator classifies retrieved context confidence to trigger refinement, web search, or direct generation.
  • Figure 4: Performance Comparison on Higress Blog Datasets. The optimized system achieved >90% recall, significantly outperforming the naive baseline in retrieving relevant community knowledge.
  • Figure 5: Performance Comparison on Higress Official Docs Datasets. The integration of Boost Ranker effectively filtered outdated versions, boosting the factuality score for configuration-centric queries.