LawMate : A Locally Deployable RAG System for Accessible Indian Legal Assistance Using TinyLlama and FAISS

@article{1718611,
      author = {Shreya Patil, Pratik Nakashe, Aditya Medhekar, Sourabh Sadake, Jyoti Bansode},
      title = {LawMate : A Locally Deployable RAG System for Accessible Indian Legal Assistance Using TinyLlama and FAISS},
      journal = {Iconic Research And Engineering Journals},
      year = {2026},
      volume = {9},
      number = {12},
      pages = {391-400},
      issn = {2456-8880},
      url = {https://www.irejournals.com/formatedpaper/1718611.pdf},
      abstract = {Access to Indian legal information remains limited for non-lawyers due to complex statutory language, fragmented resources, and high consultation costs. This paper presents LawMate, a lightweight Retrieval-Augmented Generation (RAG)–based legal assistant designed to provide citation-grounded responses for everyday legal queries. The system integrates TinyLlama (1.1B parameters) with FAISS-based se-mantic retrieval over official Indian legal documents, ensuring responses are generated strictly from retrieved context to reduce hallucination. A FastAPI backend and React-based frontend enable efficient local deployment on resource-constrained devices. Experimental evaluation demonstrates high retrieval relevance, low response latency, and practical usability for citizen-facing applications. The proposed architecture provides a scalable, low-resource solution for transparent and accessible legal assistance in the Indian context.},
      keywords = {Legal AI, Retrieval-Augmented Generation, TinyLlama, FAISS, Indian Law, Offline Chatbot, Semantic Search, FastAPI.},
      month = {June},
      doi = {https://doi.org/10.64388/IREV9I12-1718611}
  }