APEX AI — MULTI-MODEL HYBRID CHATBOT ASSISTANT FOR HYBRID ONLINE AND OFFLINE ENVIRONMENTS

APEX AI is a full-stack, multi-model hybrid chatbot assistant designed to deliver intelligent conversational
capabilities in both online and offline environments. Built on a FastAPI backend and a single-page responsive
frontend, the system integrates four large language models — Groq Llama 3.3 70B, Google Gemini Flash 1.5,
Ollama Phi-3, and Ollama TinyLlama — providing seamless automatic switching between cloud-based and
locally hosted inference engines depending on internet availability. The platform supports multimodal inputs
including text, images (via Gemini vision), and file uploads, real-time DuckDuckGo web search, and browsernative voice input. Session state is managed entirely in-memory using Python dictionaries, ensuring zero-database
overhead and rapid deployment. Experimental evaluation confirms sub-500 ms latency for online models and
graceful offline degradation without user-visible interruption, making APEX AI a practical solution for
connectivity-constrained environments.