vitna.ai/docs/cli/aperture/seed-notify Machine view
Raw .md llms.txt
> Source: https://vitna.ai/docs/cli/aperture/seed-notify
> Raw Markdown: https://vitna.ai/docs/cli/aperture/seed-notify.md
> Corpus index: https://vitna.ai/llms.txt

# Seeding and notification

Loading the engine's reference data, and draining the outbox into the feed that
stands in for email.

```bash
npm run aperture:seed
npm run aperture:notify
```

## `aperture:seed`

The engine's reference data, idempotent: the org, the first workload, its
candidates, its evals, and the six decisions of its history.

Those six are human decisions taken before the engine existed. They are seeded
as decisions rather than reconstructed as proposals, because they are what the
workload actually did, and a register showing only the machine half would
misreport its own history.

## `aperture:notify`

Drains the outbox.

Notification is a feed rather than email, and the reasoning is recorded in the
source: email needs SMTP details nobody has, an inbox nobody has agreed to
monitor, and a deliverability problem this project does not need. A feed needs
no credential, and every consumer already knows how to read one.

So the published channel is [`/api/v1/feed.json`](../../api/v1/log/feed.md), and email
stays documented as an add-on for whenever there is an inbox somebody reads.

## What a notification says it did not do

A fired trigger's feed item states that nothing was routed and nothing was
accepted.

That sentence is in the payload rather than only in documentation, because a
notification implying the system took an action it did not take is the worst
thing this channel could carry.
Human Machine Shift+M