DocuSearch is an AI-powered document search and retrieval system that allows users to efficiently search for and retrieve relevant document chunks based on their queries. The system leverages pre-trained NLP models and similarity search techniques to provide accurate and fast retrieval of information from a large collection of documents.
User-friendly web interface for submitting document retrieval queries. AI-powered processing of user queries using HuggingFace Transformers. Efficient document retrieval using Milvus, an open-source vector database. Support for chunking large documents to optimize retrieval performance. Scalable deployment with GPU acceleration on Nutanix Kubernetes Engine (NKE) Cluster.
Frontend: Next Js, Axios Backend API: Flask, Hugging Face Transformers, PyTorch Document Retrieval Engine: Milvus, Python SDK for Milvus Container Orchestration: Kubernetes, Docker GPU Acceleration: Nvidia P40 GPUs Database: PostgreSQL or MongoDB Load Balancer: Nginx or HAProxy Cloud Platform: Nutanix Cloud Platform (NCP), AWS/GCP/Azure (optional) Monitoring and Logging: Prometheus, Grafana, ELK Stack Testing: PyTest or unit test Documentation: Sphinx or MkDocs
Docker is installed on the system for containerization. NVIDIA GPU drivers and CUDA toolkit for GPU acceleration (if using Nvidia P40 GPUs).