Build an agent that answers from your own documents

Helps a person build an assistant that looks things up in their documents before it answers. It loads the documents into a searchable index, tags the pieces, and tests with real questions.

What it asks for

  • about (Optional, words): What this run is for, in the person's words.
  • known (Optional, words): What is already known, decided or done, so the run starts from it instead of asking again.

The steps your AI walks

  1. Put the documents in one place and name who will ask what
  2. Load the documents into an index that finds pieces by meaning
  3. Tag each chunk so answers can be narrowed
  4. Build the agent that looks things up before it answers
  5. Test with the written questions and fix by reading the failures
  6. Keep the index growing and record what is left

Use it from your AI

Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this workflow by what it does and runs it.

Say to your AI: “Use Trillion to helps a person build an assistant that looks things up in their documents before it answers. It loads the documents into a searchable index, tags the pieces, and tests with real questions”