RAG Manuals Tech Support

The Problem

Tech support staff spent significant time digging through arcade equipment manuals for every ticket. Generic PDF search finds keywords but can't answer questions. Manuals vary wildly in format — tables, flowcharts, buried paragraphs.

The Solution

A custom RAG web app over the full library of equipment manuals. User picks the specific game, asks a natural-language question, and gets three coordinated outputs: a synthesized answer, a PNG snapshot of the actual manual page, and a link to the public manufacturer manual.

Key Features

• Per-game metadata-scoped retrieval (collapses search to one machine's docs)
• Layout-aware PDF ingestion (preserves tables, diagrams, multi-column)
• Visual-proof responses: answer + page PNG + public manual link
• Source-grounded answers techs can verify with one click

Tech Stack

Outcome

Technically sound system with the trust-building visual-proof pattern. Key lesson: the page PNG snapshot is what makes it trustable by field techs — they don't trust answers they can't verify.