Mochi: Prompt-to-Production Website Generation Using Next.js, OpenAI GPT Models, Inngest Orchestration, and E2B Sandboxing

Website development traditionally demands extensive technical expertise, significant time investment, and manual coding effort, creating substantial barriers for non-technical users, small businesses, and rapid prototyping scenarios. Recent advancements in large language models have opened new possibilities for automating complex development workflows through natural language understanding. This paper presents Mochi, an intelligent website builder that achieves prompt-to-production website generation using Next.js, OpenAI GPT models, Inngest orchestration, and E2B sandboxing. The proposed system enables users to describe website requirements in plain natural language, which are then automatically translated into complete, functional, and responsive web applications. The architecture integrates four key components: a Next.js-based frontend for server-side rendering, OpenAI API for natural language processing and code generation, Inngest for event-driven workflow orchestration, and E2B sandbox environments for secure isolated code execution and validation. Users interact with an intuitive interface where textual prompts are processed through the OpenAI language model to extract intent, generate appropriate layouts, components, and styling, which are then validated in sandboxed environments before real-time rendering. Experimental evaluation using diverse website requirement prompts demonstrates that Mochi reduces development time by approximately 85% compared to traditional manual coding, achieves 92% accuracy in intent interpretation for well-structured prompts, and generates production-ready websites within an average of 45 seconds. The system successfully handles various website categories including portfolios, business landing pages, and content-driven sites while maintaining responsive design principles and modern web standards. Results validate that AI-driven automation combined with secure execution environments can democratize web development, enabling individuals without programming knowledge to create professional websites efficiently. This work contributes to the growing field of AI-assisted software development and demonstrates practical implementation of large language models in automated code generation workflows.