This study evaluates the effectiveness of Sarvam-105B India's flagship sovereign large language model, in complex library operations. The framework is based on our prior research. The framework includes OpenRefine as a front-end data transformation tool and its AI extension which is configured by using API key, model name and chat completion URL of Sarvam AI. The contextual precision of Sarvam-105B demonstrated across five core library operations: genderization of author names, translation of technical abstract, summarization of abstract, sentiment analysis, multilingual bibliographic record generation. Because of the extensive training in Indic corpora Sarvam-105B has higher linguistic precision, especially in the major Indian languages. In methodology, the 'connected' approach like transliteration of names before genderization or summarization from translated Bengali abstract shows that domestic AI is capable of performing complex, multi-layer tasks. The research findings reveal that sovereign LLMs are powerful enough to fulfill the needs of professional libraries and also maintain linguistic sovereignty, cultural integrity and data sovereignty. The study provides a roadmap for implementing sovereign AI stack into the libraries that prioritize regional accuracy compared to foreign proprietary LLMs.
Neogi et al. (Mon,) studied this question.