🔗 GitHub Repository

🔗 Demo

Description

Shakti is an intelligent Retrieval-Augmented Generation (RAG) chatbot designed to assist substation workers with queries related to equipment maintenance, testing procedures, safety measures, and troubleshooting steps. It was developed as a solution for the SIH1380 problem statement in the Smart India Hackathon, focusing on providing an AI-driven assistant for substation asset maintenance.

The chatbot is capable of answering questions about maintenance activities for various substation equipment, including Transformers, Reactors, Circuit Breakers, Instrument Transformers, and Surge Arrestors. By leveraging semantic search and natural language processing (NLP), it retrieves relevant information from documented procedures, industrial standards, and safety guidelines to provide accurate and context-aware responses.

Key Features

  • RAG-Based Query Processing: Combines retrieval-based search with a language model to generate precise answers.
  • Equipment-Specific Guidance: Provides step-by-step instructions for maintenance tests and checks.
  • Standard Compliance: Integrates industrial safety standards and recommended test limits.
  • Troubleshooting Support: Suggests actions to resolve issues encountered during maintenance.
  • Scalable & Secure: Built using Django and LangChain, with ChromaDB for efficient vector search.

Tech Stack

  • Large Language Model: Mistral 7B Instruct v0.1
  • Embeddings: BAAI/bge large english
  • Language Model Integration: LangChain
  • Vector Database: ChromaDB
  • Backend: Django (Python)
  • Frontend: Bootstrap

Shakti enables substation personnel to access critical maintenance information instantly, improving efficiency and safety in power infrastructure operations.