Skip to content Skip to content

Loading your own documents into d3kOS for AI-assisted reference

Viewing 1 post (of 1 total)
  • Author
    Posts
  • #482
    Skipper Don
    Keymaster

    The d3kOS RAG (Retrieval-Augmented Generation) system lets you load your own documents — engine manuals, safety procedures, equipment manuals — so Skipper Don can answer questions from them. Here is how it works and what formats are supported.

    What RAG does. RAG splits your documents into small chunks and creates a searchable index using embeddings (a mathematical representation of meaning). When you ask Skipper Don a question, d3kOS searches the index for relevant chunks and provides them to the AI model along with your question. The AI answers using your actual documents rather than general knowledge.

    Supported formats. PDF and plain text files. PDF is the most common format for engine manuals and safety documentation. The d3kOS ingestion tool (accessible via the Admin panel) accepts PDF uploads directly.

    What to load. Your engine manufacturer’s manual is the most valuable document to load — it contains fault codes, maintenance procedures, torque specifications, and troubleshooting flowcharts specific to your engine. Safety authority publications (Transport Canada Safe Boating Guide, USCG Navigation Rules) are also useful. Any document you consult regularly is a candidate.

    Limitations. The RAG system searches for relevant text but does not reason across the entire document simultaneously. Very large tables, diagrams, and wiring schematics do not translate well from PDF to searchable text. The system works best with text-heavy documents like manuals and regulations.

Viewing 1 post (of 1 total)
  • You must be logged in to reply to this topic.

Comments are closed.