RevyAI Multi-Agent System

RevyAI Multi-Agent System — a multi-agent AI system built on LangGraph designed to automate client communication for RevyAI, a business AI automation company.

Goal: Handle inbound client interactions intelligently by routing conversations to the right agent based on intent.

Agents:

Supervisor — analyzes user message, detects intent, refines search query, and routes to the appropriate agent

Direct Node — handles greetings and general conversation without RAG

Sales & CS Node — answers questions about services, pricing, and company info using RAG retrieval from a knowledge base

HR Node — handles job inquiries and hiring-related questions

Tech Stack:

LangGraph — agent orchestration

Ollama / OpenRouter — LLM inference

ChromaDB — vector database

Sentence Transformers — embeddings

LangChain — tools and prompt management

Flow:

User Message

↓

Supervisor (intent detection + query refinement + routing)

↓

Direct Node / Sales&CS Node / HR Node

↓

ResponseSonnet 4.6