Agentic AI

What: An explicit refill graph: intake, policy, script, safety, decide, review, act. Local tools pick the path. The model writes the note.
Why: A hidden refill bot is hard to trust. A visible graph — and a pharmacist click on escalate — is the lesson.
What you are seeing: Click a sample prompt to read what that path means in plain language, then watch Flow on the left. AWS Architecture on the right is that step’s box — click the title for the stacked twin. Controlled, expired, and labs pause for Approve or Deny. This is a fictional teaching chart, not a clinic.

Static GitHub Pages demo — sample questions replay recorded runs with paced animations. Vector DB hashes in this tab and does not call localhost. Clone the repo and make dev for live labs.

What this prompt does

Patient A — the easy refill

Patient A asks for lisinopril, a common blood-pressure pill. Name, date of birth, and zip match the teaching chart. The prescription still has refills. The last fill was long enough ago. The drug is not tightly controlled, and no lab checks are overdue.

What happens. The system fills it on its own. No pharmacist has to click. This is the boring happy path — everything checks out.

Flow

AWS Architecture

START
accept the request
END
return the outcome

Node output

Run a sample request. Nodes light as each step finishes. Click a finished node to inspect it.