JevMade Sign in
← Back to guides

JevMade field notes / Integration walkthrough

Help an AI assistant find the right saved information

This guide shows how to help an AI assistant find useful saved facts. It uses a tool to ask yes-or-no questions about many facts at once, keeping only the information needed to answer a user.

Original by samdotmakRetrieval

Listen to this guide

JevMade’s plain-English explanation

0:00 /

Our summary

This project helps an AI assistant find the right saved facts to answer a user. Normal search methods often miss important details if the words do not match exactly. This approach tries to find useful information without being too slow or expensive.

The software uses Jev, an AI tool that chooses from options rather than writing an answer. It gives each saved fact a short label and asks Jev a yes-or-no question about it. The software then keeps only the facts that score high enough to pass a chosen cutoff point.

This is for people building AI software. To use it live, you need a private access code from TypeSafe. You must test your cutoff point with real examples to ensure the software does not miss important facts or keep useless text.

Key takeaways

  1. Give each saved fact a label so the software can organize answers without generating new text.
  2. Group yes-or-no questions together in one request to save time when checking many different facts.
  3. Test your cutoff score with examples of missed facts and useless text to find the right balance.

Live use requires a TypeSafe private access code. The speed, cost, and performance numbers are the author's own tests, not independent guarantees.

GitHub README and core module

Read the original guide Opens the author’s site in a new tab.

Keep this for later

Sign in to bookmark experiments, guides and videos, and keep notes only you can see.

Continue to sign in

We’ll bring you back to this listing.