Architecture

A secure, monetised data pipe — nothing more.

We are not building a chat interface. Your AI assistant handles the conversation and reads the documents; MINDSET is a serverless API endpoint that authenticates the user, enforces usage quotas, queries the vector database, and returns a structured JSON payload — which your AI assistant then turns into a clear, ready-to-use answer.

The request pipeline

@MINDSET call

from AI agent

Authenticate

OAuth 2.0 · 2 devices

Vector search

finds the closest matching models

JSON payload

your AI assistant turns it into a clear answer

Clean, queryable schema

A 107k-word book and a 60k-word reference appendix are parsed, fuzzy-matched on topic name, merged, and embedded into a single normalised record per model.

topic record

{
  "topic_id": 148,
  "topic_name": "Peak–End Rule",
  "summary": "People judge an experience…",
  "brand_examples": ["IKEA", "Uber"],
  "industry_tag": "Retail"
}