You stand at the seed rack in March and try to remember whether the sungold you grew last year was the one that did well or the one that split in the heat. You know it was one of them. You probably noted it somewhere — a half-filled garden journal, a photo with caption, a text to your sister where you bragged or complained. By the time you actually need the information, it's spread across so many places that you just buy the variety and hope.

Most gardeners run into this version of the same problem. The information exists across seed packets, photos, weather memories, and the part of your brain that's good at remembering people's names but not whether last spring was an unusually warm one. Each season starts roughly from scratch.

A vault that holds one parent page per garden — with planting dates, varieties, harvests, soil notes, and the year-over-year history — fixes most of it. The agent does the cross-season comparison. (For the related hobby pattern across observation logs more generally, [Notes for Hobbyists: From Woodworking to Astronomy in One Vault](/guides/personal-life/ai-notes-hobbyists-woodworking-astronomy/) is the broader companion.)

## One parent page per garden, with the seasons underneath

In Docapybara, the garden gets a parent page. Pages nest with no depth limit, OneNote-style, so a parent page called *Garden* can have child pages for *Beds*, *Varieties tried*, *Planting log*, *Harvests*, *Pests & disease*, *Soil*, and *Seasons*.

For larger gardens (or multiple plots), each bed or zone can be its own child page. The agent treats the whole nested structure as one searchable pile.

The single most useful structure to start with is *Seasons* as a parent for the year-over-year archive. *2026 spring*, *2025 fall*, *2024 winter* — each as a child page that holds what got planted, what came in, what went wrong, and what you learned. Two years of these and you have a real reference; ten and you have something close to local expertise about your specific microclimate.

## A planting log with dates and varieties

The single most useful page is *Planting log*. Date, what got planted, the specific variety (not just "tomato"), where it went, the source of the seed or start, and any conditions worth noting (was it a cold spring? did you start late?).

An inline database does this well at scale. The `:::database:::` directive lets a small spreadsheet live alongside whatever prose context you want. Columns for date, crop, variety, location, source, and notes.

The agent can update it. *"Add today's planting — six sungold tomatoes from the local nursery, two Cherokee Purple from seeds I started in February, all in the south bed."* Row appears with the date.

After a season or two, the agent can pull patterns. *"Compare this year's tomato planting to last year — what changed, what stayed the same."* Or: *"What was the latest I've ever started peas and had them work?"* The answers come back grounded in your real log, not your impression of what you usually do.

## Varieties tried — the page that earns its keep over years

For gardeners who care about which specific variety to grow next year, a *Varieties tried* page (or database) is the highest-leverage thing in the vault. Every variety you've grown, with notes on how it performed in your specific conditions.

Columns for variety, type (cherry tomato / pole bean / etc.), years grown, performance (rough rating), notes on flavor, vigor, disease resistance, and whether you'd grow it again.

The seed catalogs all sound the same in February. The difference between "great in our garden" and "a disappointment in our garden" comes from your own notes — and almost no one keeps them in a way that's actually retrievable.

The agent makes this useful. *"What cherry tomatoes have I grown that I rated highly, and which ones do I keep coming back to?"* The answer is grounded in your own data. *"Suggest three new varieties to try this year that fit conditions where things like [variety I liked] did well."* The agent's `web_search` tool can pull current reviews and seed sources for the suggestions.

## Harvests so the season has a real shape

A *Harvests* log is its own page. Date, crop, rough quantity (a basket, a pound, a bowl), and any notes. You don't need to weigh things. Rough is better than nothing.

Voice is often the right capture tool here, because you're outside, your hands are dirty, and you're not going to type. Tap record, talk for fifteen seconds. *"Tonight pulled the first basket of green beans, second wave of zucchini, the cucumbers are starting to slow down."* The transcript lands on the harvest log with a timestamp.

After the season, ask the agent. *"What did the season actually look like — what came in heavily, what came in lightly, when did each crop peak?"* You get a real summary, not a guess. The next year, you plan with last year's actual yields in front of you.

For gardeners who preserve, can, or share with a CSA-style network, a harvests log is also the substrate for planning next year's planting volumes. *"We froze a lot of pesto last year and ran out in February — how much basil did we actually grow, and how much would we need to make pesto last through March?"* The math is doable if the data is there. (For the kitchen side that catches your harvest, [Documenting Recipes and Cooking Experiments Without Losing the Plot](/guides/personal-life/document-recipes-cooking-experiments/) covers it.)

## Soil notes, amendments, and what's been done where

Soil is the long game. Whatever you do to a bed this year shows up in next year's results. A *Soil* page (or per-bed soil notes) holds the amendments, the test results if you've ever done one, the cover crops, the compost additions, and any rotation plan you're trying to follow.

For test results, drop the lab PDF on the page. PDFs upload and convert to markdown automatically, so the actual numbers — pH, NPK, organic matter, micronutrients — become searchable text. *"What was the pH of the south bed last year, and have I added anything that would have moved it?"* The answer comes back grounded in the actual report.

Crop rotation is one of those gardener things that's hard to track without notes. A simple per-bed history — *south bed: 2024 tomatoes, 2025 garlic + brassicas, 2026 ?* — and a one-sentence rotation rule (*"don't put nightshades back for three years"*) lets the agent help with planning. *"Given the bed history and my rotation rule, where should I put tomatoes this year?"* The answer is grounded in your real history.

For gardeners doing more involved work — biochar, mycorrhizal inoculants, fermented plant juice, specific microbial inputs — the soil page becomes a research log too. Drop the articles you're learning from, paste the YouTube transcripts of the sources you trust, and the agent can quote them back when you're trying to remember why you're doing what you're doing.

## Pests, disease, and the things that went wrong

Every garden has a year where something went wrong — late blight, vine borers, an unexpected hailstorm, a deer family that found the bean trellis. A *Pests & disease* page is where these get recorded, in the moment, with the date and the response.

Photos are useful here. Drop the picture of the leaf with the spots, write a one-line caption, ask the agent to help identify it. *"This was on my zucchini today — what is it likely to be, and what's the standard organic response?"* The agent can use `web_search` for current sources and ground the answer in what you uploaded plus what's in your existing notes.

Across years, the pattern matters. *"I had cucumber beetles three of the last four years — what timing did they show up, and what did I try?"* The historical answer is the basis for next year's plan, instead of starting fresh every spring.

For gardeners using IPM (integrated pest management) or any deliberate strategy, the page can hold the actual strategy and the agent can apply it. *"My strategy says spray neem only on weekly intervals when beetles cross threshold — given my recent observations, what should I do this week?"* The reasoning is in your own framework, not a generic gardening blog.

## A garden plan and the seed order that follows from it

Each season starts with a plan. A *2026 plan* child page (or a *Spring 2026* page if you do multiple plantings per year) holds the rough layout — what's going where, in what quantities, with what dates.

The agent can build a draft. *"Looking at my varieties tried, my soil page, my bed history, and my harvests last year — draft me a plan for spring 2026 across the four beds, with rough planting dates for my zone."* You'll edit it. But editing is much faster than building from zero.

The seed order then follows from the plan. The agent can make the list. *"Look at my draft plan, cross-reference with the seeds I still have on hand from last year (on the *Seeds inventory* page), and tell me what to actually order."* The list comes back; you order; the seeds-inventory page gets updated when they arrive.

A complementary habit some gardeners find helpful: a *Lessons* page at the end of every season. *"What worked, what didn't, what I'd do differently next year."* Five minutes of voice notes is enough. Three years of these stacked up turn into the most useful gardening advice you'll ever own — because it's specific to your conditions. (Retired gardeners often pair this with [Notes for Retirees](/guides/personal-life/ai-notes-for-retirees/) for the broader project shape.)

## A starter shape that works on day one

If you're moving from "everything in my head and a few seed packets" to a vault, this is what we'd suggest:

- **A *Garden* parent page** — bed layout sketched in prose, your zone, the basics.
- **A *Planting log* database** — date, crop, variety, location, source. The single highest-leverage page.
- **A *Varieties tried* database** — earns its keep starting in year two.
- **A *Harvests* log** — voice notes work fine.
- **A *Seasons* parent** — one child per season, with a *Lessons* summary at the end.

That's it. No need for color coding or templates. The vault grows the way the garden does — slowly, across seasons, with a real history accumulating underneath.

The point isn't to turn gardening into a database project. It's that the small amount of structure you keep means next year's planning starts with last year's actual data, the variety you loved is findable by name, and the agent can help with the cross-year comparisons that are otherwise impossible to do from memory.

[Try Docapybara free](/accounts/signup/) — start with a planting log and a harvests page this season, and a year from now you'll have something most gardeners wish they'd kept.