AI database for production equipment setup and repair: vector knowledge base + Telegram bot

I received an unusual order: a database of equipment settings and repairs. Honestly, I didn't even know they did that. But a customer from Germany convinced me: their competitors are actively using AI-powered databases of settings and repairs in production.

What we did

We compiled a vector database of all the company's equipment. Sources: company manuals, YouTube training videos, and two specialized services:

• iFixit is a community and database of millions of step-by-step repair and disassembly guides, complete with photos and tips.
• ManualsLib is one of the largest free libraries of original manuals, datasheets, and schematics in PDF format.

The interface is a user-friendly Telegram bot on the QuestFlow service: the installer types or speaks a question, attaches a photo, and receives a response with steps.

How does this work

The AI understands literally everything: from mundane basic settings ("I placed the workpiece, applied a ruler—what next?") to complex breakdown scenarios. The response is based on the company's specific documentation, not on general knowledge of the model, so there are no "hallucinations" with other machine models.

Who needs this?

For those who want to digitize the knowledge of all experienced adjusters and combine it into a single directory: so that a new employee can work like a veteran within a week, and the knowledge doesn't disappear along with the rest of the team.