I Built an AI Receptionist for a Luxury Mechanic Shop – Part 1

AI answers the phone, wallets cheer, ‘luxury’ purists boo

TLDR: A mechanic’s brother built “Axle,” an AI receptionist to stop missed calls and keep pricey jobs from slipping away. Commenters split: luxury-lovers want a human voice, pragmatists love the savings, and a side debate rages over whether the fancy tech is overkill or just good guardrails.

A luxury mechanic’s brother built “Axle,” a phone-ready AI receptionist that knows prices, hours, and policies, and only answers from the shop’s own info. The goal: stop losing jobs when no one picks up. It even uses a safety trick called RAG (the bot looks up answers instead of guessing) so it won’t quote $200 for a $450 brake job. Cool, right? Cue the comment brawl.

The loudest reaction: “Luxury means a human.” Critics say the word “luxury” and a robot voice don’t mix. One summed it up as, if you’ve reached the phone, you need nuance—not a website read-back. Another dunked with the meme-y line “clanker != luxury.” Meanwhile, a practical camp shrugged: this will save money and most callers won’t care—as one put it, their parents wouldn’t notice or mind if the bot just books them.

Nerd fight of the day: Is RAG overkill? Some say a few prices and hours don’t need fancy lookups—just paste it all in. Others argue the structure helps avoid “oops” quotes and keeps answers tidy.

Bottom line: DIY family fix meets vibe police. Fans cheer the hustle and faster pickups; haters insist “luxury” means a real person. Either way, Axle just rolled onto the phone line—and the comments are revving.

Key Points

  • The project builds a custom AI voice receptionist (“Axle”) to answer a mechanic shop’s calls and reduce lost revenue.
  • A RAG pipeline powers accurate answers using a knowledge base of 21+ documents scraped from the shop’s website.
  • Embeddings are created with Voyage AI (voyage-3-large) and stored in MongoDB Atlas with Atlas Vector Search for retrieval.
  • Anthropic’s Claude (claude-sonnet-4-6) generates responses constrained by a strict system prompt to avoid hallucinations.
  • Telephony integration uses Vapi with Deepgram (STT), ElevenLabs (TTS), a FastAPI webhook, and Ngrok for development exposure.

Hottest takes

“if I get an LLM receptionist, that ain’t it” — NiloCK
“Is RAG even necessary here?” — pbmonster
“my mom or dad would most likely not notice or care” — yuppiepuppie
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.