Smart Inventory CRUD Web Application with NLP

Abstract

Small-to-medium enterprises operating warehouse and retail environments continue to rely on igid, form-driven, single-language inventory tools, thereby incurring elevated data-entry error rates, negligible analytical depth, and systemic exclusion of non-English-speaking operators. This paper presents the design, implementation, and rigorous empirical evaluation of an AI-augmented Smart Inventory CRUD Web Application built on the MERN stack (MongoDB, Express.js, React.js, Node.js). Unlike prior work that treats NLP-based querying and structured CRUD management as separate concerns, the proposed system integrates both interaction modalities within a single deployable platform. A cloud inference pipeline anchored by the GROQ API and the LLaMA-3 70B language model translates free-form operator utterances—entered via keyboard or voice—into validated database transactions spanning the full Create-Read-Update-Delete lifecycle. A companion AI Report Generator aggregates multi-dimensional inventory telemetry and synthesises structured intelligence reports encompassing key performance indicators, stock health diagnostics, demand-trend projections, overstock and stockout risk estimates, and data-driven replenishment recommendations. A browser-native voice interface extends these capabilities to English, Hindi, and Marathi speakers without server-side audio processing. Systematic evaluation across 100 functional test cases, a JMeter concurrent load simulation, a 200-utterance multilingual speech corpus, and structured user acceptance trials with 15 warehouse operators yielded: 90.4% NLP command accuracy, 142 ms mean CRUD response latency, 3.2 s average report synthesis time, 91.7% English voice recognition accuracy, 86.2% combined Hindi/Marathi voice accuracy, and 99.6% system uptime under more than 1,000 concurrent sessions—meeting or surpassing every specified performance target. Comparative evaluation confirms that the proposed system is the only reviewed solution satisfying all six evaluated capability criteria.