Test the Conversations server

Use the prompts below to confirm the Conversations server is working and to see how your AI assistant handles different shopper questions. Each prompt is something you can send to your AI assistant directly. All examples assume you're connected to the Pacific Apparel sample catalog — see Connect the Conversations server.

Verify the connection

Send your AI assistant a basic catalog question:

What products are in this catalog?

Your AI assistant should describe the Pacific Apparel catalog or return example products — jeans, tops, shoes, or accessories from brands like Roadster, Pacific, Levis, or Adidas. If it responds with a generic answer like "I don't have access to a catalog," check your configuration in Connect the Conversations server.

Specific-product searches

These prompts simulate shoppers who know what they want. Your AI assistant should call the search_products tool and return ranked results from the catalog.

I need slim fit jeans for men, under $40.

Women's running shoes, size 8, under $120.

Show me wedding bands over $100.

Show me casual shirts from Roadster.

Anything on sale for Valentine's Day?

When strict filters return too few results, your AI assistant relaxes the least important filter and tells the shopper what it changed. This prevents the conversation from hitting a dead end.

Open-ended discovery

These prompts simulate shoppers who are browsing or unsure of what they want. Your AI assistant should call the seeker_products tool to ask follow-up questions, then search once it has enough context.

I need a birthday gift for my husband.

I'm going on a beach holiday and need some new things.

I'm just looking around. What's popular?

Your AI assistant switches from discovery to search automatically once it has a product type, occasion, budget, or style preference to work with.

Full conversation flow

Test a complete shopping conversation that moves through discovery, search, and refinement in a single thread:

Help me find a gift for my dad. He likes cycling and my budget is around $100.

A working setup completes this in three to four turns: clarifying questions, a catalog search, then refinement based on the shopper's reaction.

What good results look like

When the Conversations server is working well, you should see:

• Products returned from the catalog with titles, prices, and other attributes.
• Filters in the shopper's message applied accurately (price, color, size, category, brand).
• Follow-up questions that match the shopper's tone and surface relevant catalog attributes.
• Mid-conversation refinements that update the previous result set instead of starting fresh.
• Smart filter suggestions that reflect what's actually in the catalog — for example, real price brackets and available colors, not generic ranges.

If results don't match these expectations, it's most often a sign that your AI assistant is misinterpreting filters or calling the wrong tool. See Best practices for tips on prompt design and prototype evaluation.