lamp.
ResourcesPricingAbout

Retrieval augmented generation

Retrieval augmented generation is a technique in which relevant text is searched out of your own documents and given to an AI model to base its answer on.

What it means

The technique has two parts. Retrieval searches a collection of your documents, such as manuals, contracts and policies, for passages that match the question. Generation then has the language model write an answer using only those passages. The benefit is that you can update documents without retraining anything, and you can ask the system to show where each statement came from. Poor search or stale files lead to poor answers.

Example

A machine-tool dealer in Rajkot uploads service manuals for 30 models. A technician asks how often to replace a spindle belt and receives the interval with a reference to page 42 of the right manual.

Records and checks

  • Record the document list, version dates and who uploaded each.
  • Verify by test questions with known answers that the right passage is retrieved.
  • The knowledge owner reviews failed questions and updates or removes documents.

Common mistakes

  • Uploading scanned images with poor text quality, so search cannot read them.
  • Accepting an answer without opening the cited passage.
Did not find your answer?Message the lamp team on WhatsApp. Ask about setup, your records, or anything we have not covered here.Ask us on WhatsApp →