NirmataAI Website Chatbot

NirmataAI needed an intelligent, always-available assistant that could engage website visitors, answer common questions about services and pricing, and guide leads toward booking a consultation — without any human intervention.

Our team built a production-grade, streaming AI chatbot directly into the NirmataAI website using the Next.js App Router and the Vercel AI SDK. The chatbot is powered by a dynamically generated system prompt that is constructed at runtime from the company's own data files — services, FAQs, and contact information — making it a living assistant that stays perfectly in sync with the site content.

For maximum reliability, the system implements a multi-model fallback strategy: Google Gemini 2.0 Flash is the primary model, with Groq's Llama 3.3 70B serving as an automatic, zero-downtime fallback in case of provider unavailability. The entire response is streamed to the browser using the native Fetch ReadableStream API, delivering a smooth real-time typing effect.

The widget itself was engineered from the ground up as a self-contained React component — no third-party chat SDK dependencies. It features a premium pill-shaped floating trigger with an animated 'online' pulse indicator, quick-reply suggestion chips, full AbortController-based request lifecycle management, and comprehensive ARIA accessibility attributes. The result is a conversion-focused, enterprise-quality chatbot that operates entirely within the existing tech stack at zero additional infrastructure cost.