🍂 @smartplant/season

Bends the comfortable bands toward the time of year, and fades out as the plant learns its own.

A plugin for smartplant — a bridge between AI and plants.

Install

npm install smartplant @smartplant/season

Use

import { createPlant } from 'smartplant'
import season from '@smartplant/season'

const plant = await createPlant( {
  name       : 'Ivy',
  species    : 'Ficus lyrata',
  sensor     : 'mock',
  ai         : { provider : 'ollama' },
  hemisphere : 'north',   // 'north' | 'south' | 'tropical'
} )

await plant.use( season )
console.log( ( await plant.plugin( 'season' ).ranges() ).why )

The hemisphere is not guessed

Half the guesses would be exactly six months wrong, and that produces advice to cut watering back in somebody's spring — which looks like a plant doing badly rather than like a mistake. It is not inferred from a clock or a timezone. No hemisphere, no seasonal adjustment.

tropical is an honest answer rather than a third season table: near the equator there is a wet season and a dry one, and which months those are depends on the specific place. The plant's own record is the only useful guide there, and it is a better one than any table.

The table fades

It exists so a plant with two months of history is not silent, and it is blended by weight rather than applied whole — a plant halfway through its second year gets half the table and half itself. After two years it is gone: this plant's own record describes its own flat, its own window and its own radiator better than any archetype average.

API

now()

{ season, known, why }.

ranges()

{ ranges, applied, season, weight, why } — the archetype's comfortable bands, bent toward the season and scaled by how much history the plant has.

lastYear( metric = 'soil' )

What this plant itself did at this point in the year, once it has been through one. Not an average of the whole history — an average of this month last year, which is the only part of the record that describes this part of the year.

weight()

{ weight, why } — how much the table still counts for.

Test

npm test

Runs against the mock sensor and mock AI provider — no hardware, no API key, no network.

📜 License

MIT — see LICENSE.