Shipped in 5 daysAI agents + LLM function-calling (structured data search)
AI Inventory Assistant with Real Function-Calling
An agent that actually calls a search tool over live inventory — not a scripted FAQ bot.
live · interactiveFull screen
The client
A multi-brand car dealership with a messy, real-world inventory feed
The problem
Staff spent hours a day answering "do you have anything blue" by scrolling a spreadsheet where the same blue shows up as a dozen different factory paint names across brands. The client wanted customers to just ask, in plain language, and get a real answer from live stock.
What I built
- LLM agent with a real search_inventory function tool — not string-matched FAQs
- Color-synonym resolution across dozens of real factory paint names per hue
- Brand-tab and reserved-unit filtering handled inside the tool call
- Tool-call trace visible in the chat, so the logic is auditable, not a black box
- Same pattern extends to any structured catalog: parts, units, rooms, appointments
The outcome
Cut "do you have X in stock" questions from a staff task to a 10-second self-serve answer, around the clock.
Stack
Next.jsTypeScriptOpenAI function callingPostgresPrismaTailwind
More builds like this
Want something like this?
Fixed price, fixed timeline. Tell me the scope and I’ll get you a working demo fast.