Building a Knowledge Retrieval Agent

In this challenge, you will build a complete agent from scratch, configure tools, and test with real questions.

Goal

Build a Northwind Analyst agent that can answer structured queries, semantic searches, and multi-hop relationship questions.

Before You Start

Make sure you have:

  • Northwind loaded with embeddings and a vector index (Module 2, Lesson 0)

  • Tool authentication enabled for your instance

Steps

  1. Create a new agent in Aura Console → Data Services → Agents → Create Agent

  2. Configure basics: Set name to "Northwind Analyst", add a description, and connect to your Northwind instance

  3. Write instructions: Tell the agent its role and what it can do. For example: "You are a retail analyst with access to Northwind data. Answer questions about customers, products, orders, and suppliers."

  4. Add Cypher Template tools: Create tools for common queries like Get Customer, Products by Category, Top Customers by Order Count

  5. Add a Similarity Search tool: Connect it to your product_text_embeddings index

  6. Enable GraphRAG if available for community summaries

  7. Save and test with the questions below

See Aura Agent documentation for tool and GraphRAG configuration details.

Test Questions

Run these questions against your agent:

  • "Which customers have placed the most orders?"

  • "What are the top 5 products by revenue?"

  • "List products in the Beverages category"

  • "Find customers who ordered products from more than 2 different suppliers"

  • "Summarize what you know about customer ALFKI and their order history"

If you have a vector index:

  • "Find products similar to spicy condiments"

  • "Find products similar to hot sauce"

What to Check

Verify that your agent:

  • Selects the right tool for each question

  • Returns accurate results that match what you would get from a direct Cypher query

  • Handles relationship-heavy questions using GraphRAG context when enabled

Summary

In this challenge, you built a knowledge retrieval agent with Cypher tools, Similarity Search, and GraphRAG.

Chatbot

How can I help you today?