Multi-Module Collaboration: Ethnic Costume Culture Knowledge Agent System Based on Large Language Models

Large Language Models (LLMs) have advanced rapidly and are increasingly applied to specialized research domains. However, when applied to such domains, LLMs continue to face persistent challenges: knowledge base integration alone offers limited flexibility and poor timeliness; knowledge graph construction is labor-intensive and difficult to scale; and fine-tuning risks catastrophic forgetting and requires extensive computational resources. As a result, LLMs applied to specialized, data-sensitive fields such as ethnic culture still suffer from hallucination, poor timeliness, and restricted functionality. To address these limitations, this paper proposes a novel multi-module collaborative agent framework that moves beyond simple module integration. Its core innovation lies in the integration of a knowledge-based RAG module and a culturally fine-tuned image generation module within a unified agent architecture, coordinated by a practical, prompting-based question classifier; the empirical benefits of knowledge-base grounding and cultural fine-tuning are evaluated directly, while the incremental benefit of the coordination mechanism itself, relative to simpler routing strategies, remains an open question for future work. Using She ethnic costume culture as a case study, the system effectively mitigates the aforementioned limitations without altering the base model’s parameters. Experimental results demonstrate that our framework significantly improves the accuracy, relevance, and cultural authenticity of both textual and visual outputs, and provides a flexible, reproducible paradigm for applying LLMs in other data-sensitive and expertise-driven fields.