Skip to content

demo_paths.md

Latest commit

 

History

History
41 lines (30 loc) · 2.81 KB

demo_paths.md

File metadata and controls

41 lines (30 loc) · 2.81 KB
Raw

Demo Paths

Three scripted paths for the ElectroHub electronics retail demo. Each one is designed to show the difference between structured tool use and generic RAG.

Tip: Run each opening question once in Context Surfaces mode and then repeat it in Simple RAG mode to show how the tool-enabled system reaches live product, store, and order data.

Path 1: Product Fit to Local Pickup

  1. Ask: "I am looking for a MacMini to run OpenClaw on, what machines do you have in stock that would help?" Expected outcome: the agent recommends 2-3 computer products instead of TVs or accessories. The strongest fits are usually ByteForge Micro Pro, NovaBook 14 Creator, and PulseTower Mini RTX 4060, with concrete hardware and price details.
  2. Follow with: "And can I pick that up at my local store?" Expected outcome: the agent uses the signed-in profile, identifies ElectroHub Cherry Creek as the home store, and confirms same-day pickup for the recommended products that are in local inventory.
  3. Follow with: "Which one is the best value?" Expected outcome: the agent compares price, form factor, and headroom, typically positioning ByteForge Micro Pro as the best compact value and PulseTower Mini RTX 4060 as the stronger performance step-up.

RAG contrast: Simple RAG can talk generally about buying a PC for retro games, but it cannot know what is actually in stock or what Maya's local store is.

Path 2: Shipment Delay Investigation

  1. Ask: "I haven't received my shipment yet." Expected outcome: the agent finds ORD_EH_1001, reports the original promise date, current tracking number, latest scan at the Salt Lake City facility, and the weather-delay reason.
  2. Follow with: "When should I expect it now?" Expected outcome: the agent cites the updated estimated delivery from SHIP_EH_001 and distinguishes it from the missed original promise date.
  3. Follow with: "Has support already opened a case?" Expected outcome: the agent surfaces CASE_EH_001 and explains that the shipment-delay case is already open.

RAG contrast: Simple RAG can explain shipping delays in general terms, but it cannot identify Maya's order, tracking number, current location, or active case.

Path 3: Store-Aware Commerce Memory

  1. Ask: "Show me my recent ElectroHub orders." Expected outcome: the agent summarizes the signed-in customer's recent shipping and pickup orders, including order totals and fulfillment types.
  2. Follow with: "Which one did I pick up locally?" Expected outcome: the agent identifies ORD_EH_1002 and names ElectroHub Cherry Creek.
  3. Follow with: "Do you offer curbside there?" Expected outcome: the agent confirms curbside support at that store using the store record.

RAG contrast: Simple RAG can describe a generic pickup policy, but it has no access to Maya's order history or her preferred store.