Enterprise Hybrid RAG

See how an AI customer-service agent combines enterprise data, Vector RAG, a lightweight knowledge graph, and deterministic rules to resolve complex distributor returns for a fictional automotive-battery company. Ask a question below and watch exactly which knowledge the agent retrieves, and why, before it answers.

ClaudeStructured RetrievalVector RAGKnowledge GraphDeterministic Rules

Try the lab →See the architecture →How it works →Eval scenarios →

Interactive lab

Ask the Customer Service Returns Agent

Supercharged Battery Co.

Customer Service Returns Agent

Synthetic distributor data · simulated

Ask about a return, refund, warranty, or policy, or click a suggestion below.

Try asking about any distributor, invoice, or policy in the .

Grounding / Evidence

What the agent actually retrieved

Ask a question to see structured facts, the knowledge-graph path, retrieved policy text, rule checks, and the final decision.

Routing

Structured facts

Graph path

Policy text

Rule checks

Decision

Architecture

The LLM reasons. Enterprise data grounds it.

User question

↓

Intent + entity extraction (LLM)

↓

Retrieval router

↓

Structured Data

Vector RAG (TF-IDF similarity)

Knowledge Graph

↓

Deterministic rules engine

↓

LLM synthesis → grounded response

Structured retrieval provides exact facts (invoice amounts, dates, quantities). Vector RAG retrieves relevant policy language from a small knowledge base. The knowledge graph resolves which policy applies to a given distributor, tier, contract, or product. Deterministic rules compute eligibility, refund amount, and approval requirements. The LLM never overrides them. Read the full walkthrough on the How It Works page.

Elegance AI Lab. Practical learning for applied and agentic AI. Supercharged Battery Co. is a fictional company built from synthetic data for this demo. Not connected to a production system.