> Source: https://vitna.ai/docs/cli/witness/rates
> Raw Markdown: https://vitna.ai/docs/cli/witness/rates.md
> Corpus index: https://vitna.ai/llms.txt

# npm run witness:rates

Publish yesterday's [reference rates](../../concepts/reference-rate.md).

```bash
npm run witness:rates
npm run witness:rates -- --verify
npm run witness:rates -- --dry-run
```

| Flag | Meaning |
|---|---|
| `--verify` | Recompute every day ever published and report disagreement |
| `--dry-run` | Write nothing |

## The rule has no free parameters

The median list price among at least three distinct providers publishing a
confirmed price that day, in exact whole millionths.

Because the inputs are append-only and the rule has nothing to tune, an
independent implementation reading
[`/api/v1/skus/{slug}/history`](../../api/v1/prices/sku-history.md) should arrive at the
same series. `--verify` is Vitna running that check against itself.

## A day is published once

A series is never restated. A correction is a new row on a later day, which is
what makes the series usable as a benchmark rather than as an opinion with a
timestamp.
