QAI Agentic AI

Persistent chats, replayed conversation history, and per-message model and token accounting — so an agent retrieves across the same database it writes to, and every call is attributable.

What it does

Retrieval runs over the same Postgres that holds the records — no second datastore to keep in sync
Conversation history is persisted and replayed to the model as context
Per-message model and token accounting on every AI path
Tool use today is per-product, in the QAI Earth agent loop; retrieval today is lexical, and the vector columns are provisioned but not yet live

The other layers

QAI Data

One governed Postgres for records, documents and vectors

QAI Cloud

Delivery, edge and tenancy

QAI Infra

The AWS substrate, entirely in Terraform