SpeLL: An Agent for Natural Language-Driven Intelligent Spectral Modeling

Spectrum large language model (SpeLL) was developed to tackle core challenges in near-infrared (NIR) spectral data modeling─the high level of expertise and substantial workload required by researchers for method selection, implementation, and optimization, particularly in light of the growing number of spectral analysis techniques and specific application scenarios. SpeLL harnesses the transformative capabilities of large language models (LLMs) and retrieval-augmented generation (RAG) technology, integrating them through natural language interaction to transform complex spectral data modeling workflows into automated operations. The core strength of SpeLL lies in its dual RAG pathways. The Code RAG provides specialized code knowledge for spectral data analysis, enabling the LLM to generate robust and domain-specific analytical scripts that address the implementation and optimization of algorithms. The Data RAG stores historical spectral data sets rich in patterns and features, intelligently matching similar data to guide the choice of algorithms and optimize modeling tasks. By offering an end-to-end automated workflow encompassing natural language understanding, RAG-enhanced code generation and execution, and an Auto-Debug mechanism, SpeLL achieves intelligent and automated NIR spectral data modeling and analysis.