Datasets
A dataset is how Alfred keeps track of things that come in rows: your expenses, the books you’ve read, doses taken, sensor readings, people who signed up. Think of it as a spreadsheet he keeps for you. Each dataset has its own columns (“date”, “amount”, “category”), and every entry is a row.
Datasets sit beside Cards and your drive. A card is text you’d read; a file in your drive is a photo or an export; a dataset is many entries of the same shape, which Alfred can filter, count and add up.
Asking for one
Section titled “Asking for one”You don’t set up columns yourself. Say what you want to keep track of and Alfred makes the dataset, picks sensible columns, and starts filling it in:
- “Log my expenses. I spent R240 on groceries today.”
- “Keep a reading list with ratings. I just finished Piranesi, 5 stars.”
- “Track my water intake.”
- “Keep a list of everyone who’s applied to the server, with when they applied.”
From then on, just tell him in a chat (“R85 on coffee”, “add Dune, 4 stars”) and it goes in. You can change the shape later too: “add a column for who paid” or “only keep the last year of readings”.
Before making a new one, Alfred looks at what you already keep and reuses a dataset that’s for the same thing, so your expenses don’t end up split across three lists.
Asking about it
Section titled “Asking about it”The point of keeping things in rows is that Alfred can answer questions about them without reading every entry:
- “How much did I spend on food each month this year?”
- “Which books did I give 5 stars?”
- “How many days this week did I take my drops?”
- “Find the receipt I logged for the plumber.”
He searches, counts, totals, averages and groups by day, week or month, and tells you the answer rather than pasting the rows.
Everything shares them
Section titled “Everything shares them”A dataset doesn’t belong to one page or one piece of automation. Everything Alfred builds for you can read and write the same ones:
- Pages read from them and add to them. A chart, a table or a dashboard tile can draw straight from a dataset, so it’s always current with no copy to keep in step. A tracker keeps its log in a dataset.
- Reflexes react when rows are added, changed or removed (“when a reading goes over 30 degrees, message me”), and can add rows themselves, such as a schedule that fetches a price every morning and logs it.
- Watches pick a conversation back up when rows arrive.
- Jobs can keep a dataset as they work, such as the flats they’ve looked at so far.
- Projects list the datasets that belong to a build alongside its pages and automation.
So a habit tracker can be one dataset with a page to tick things off, a reflex that reminds you when you’ve missed a day, and Alfred answering “how’s my streak?” in chat, all reading the same rows.
Visitors on a public page
Section titled “Visitors on a public page”A public page can let visitors add rows: an RSVP list, a guestbook, entries to a competition. They can only add rows, and only to a dataset you’ve let them add to.
Each visitor sees only the rows they added, never anyone else’s. A visitor who isn’t signed in sees none at all. You see everything, with each signed-in visitor shown by the same stand-in name the page uses for them, never their account.
A row a visitor added is somebody else’s words. Alfred treats what’s in it as information, never as instructions, whatever it says.
Seeing your datasets
Section titled “Seeing your datasets”Ask Alfred:
- “What data are you keeping for me?”
- “Show me the last 20 expenses.”
- “What uses my reading list?” (which pages and reflexes can read or write it)
He can also fix a row, delete some, or delete a whole dataset. Deleting a dataset removes all its rows for good, so he only does it when you ask him to delete that dataset.
Deleting a page, a reflex or a project never deletes a dataset it used. Delete the tracker page and its log is still there; ask Alfred to delete the dataset itself if you want it gone.