AI ENGINEERING · INTERACTIVE DEMONSTRATION

Ask the documents. Check the answer.

Try an assistant with a real AI model and a fictional catalogue. Ask a question, inspect the source excerpts and see what happens when information is missing.

YOUR QUESTION

Ask the catalogue.

Fictional event equipment rental. Ask about the kits, prices or terms. These are not Celebriam Labs offers.

Built with Llama · Llama 3.3 · Cloudflare Workers AI

Try one of these questions, or write your own:

0/600

Your question and up to two previous questions are sent to Cloudflare Workers AI to generate the response. Use sample data; do not include personal or confidential information. Conversation stays in this page’s memory. Privacy information.

A defined source of knowledge

The assistant consults six sample documents. It has no web search, private business data or live inventory access.

Sources you can inspect

Answers include source excerpts. The system checks that quotations exist in the supplied documents. A model can still make mistakes: compare its answer with the originals.

Questions before actions

An ambiguous request should prompt a clarification. Missing facts require human review. The demo cannot confirm a booking, take payment or send messages.

How this demonstration is built

A statically rendered Next.js interface sends the question to a Cloudflare Worker. The Worker selects relevant catalogue documents, calls Llama 3.3 through Workers AI, and checks the response format and source quotations before returning the result. Request size, output length and request rates are bounded. If inference fails or the free quota is exhausted, an availability message is shown.

An assistant with your own documentation?

We can review your information sources, permissions and the tasks where an assistant could help your team.

Discuss my assistant