Architecture

instantAIguru is built on three layers: a Twilio Conversation Relay that streams voice with sub-300ms latency, a JavaScript Flow Engine (JSFE) that runs agentic actions through deterministic scripts instead of LLM-chosen tool calls (zero hallucination across 200,000+ production flows), and a Hybrid RAG system that hits 97%+ answer accuracy by combining vector and keyword retrieval with multi-vendor model orchestration and automatic hallucination detection. Every layer is documented in detail below.

Common architecture questions are answered in the Architecture & Technology FAQ.

Twilio Conversation Relay

WebSocket-based voice AI with real-time streaming, STT/TTS, and multi-layer architecture.

Visual Architecture Diagrams

Click any diagram to view full size.