Ask Pyron a question the way you would put it to a colleague — "which reports mention safety concerns" — and it looks for entries that are about what you mean, not only the ones that contain those exact words. Meaning-based matching runs alongside ordinary keyword search, so a single search covers both.
Keyword vs meaning
Search looks through your entries two ways at once. Keyword matching finds entries that contain the words you typed — a site name, an asset number, a phrase copied out of a report. It is exact: the word has to be there for the entry to match.
Meaning-based matching — sometimes called semantic search — goes further. It finds entries that are about what you asked, even when they are worded differently. A search for "safety concerns" can surface an entry that describes a near-miss or a hazard without ever using those two words, because the ideas sit close together in meaning.
You do not pick between the two. One search runs both and returns a single list, and an entry that matches both ways appears once rather than twice. The entries that hold your words lead the list, and among them the ones that are also closest in meaning come first. The entries that are close in meaning but hold none of your words follow, under Close in meaning. When you search for a number, such as a patient or asset number, and an entry holds it, the search palette shows only the entries that hold it: entries that merely look alike are left out, though the full results page still lists them under Close in meaning.
A word of four letters or more also finds the words it starts: "lock" finds "locked". A shorter word, and any number, must appear whole, so "170" never finds "1700". Search reads what an entry says, never its field names: "notes" does not find every entry with a Notes field.
Each result shows the entry's title, the line that holds your words (or, for an entry close in meaning, its opening line) and where it is filed in the tree, so you can tell the right one apart at a glance. Open a result and you land on the Entries page with that entry selected, ready to read in full.
Results only ever include entries you are allowed to open, so two people searching the same words can see different results depending on which parts of the tree they can reach. To open the search box from anywhere, see Quick search.
Files, by what they say
Search reads the documents in Files the same way. A file whose name holds your words comes first; after it come the files whose text is closest in meaning, each with the passage that matched and, for a PDF, its page ("Page 4: Gas leak: evacuate and call Ops"). It searches only the files you can open, and only a file's current version. Scanned pages and photos are read too: Pyron reads the words on them once they are checked for viruses, so a scanned induction pack is found by what it says like any typed document. A photo with no words on it is found by its name alone. A name is found by its words in any order: "site 4 map" finds Map – Site 4. The assistant and automations search Files this way too, as whoever they work for: the assistant names the document and page it answered from.
When semantic search is available
Meaning-based matching depends on your organisation's entries being indexed for meaning — a background step that prepares your data so a search can compare ideas, not only words. When that indexing is in place, your questions match entries by meaning as well as by keyword.
If your organisation has not set up that indexing, keyword search still works as usual: you can find an entry by the words it contains. Only the meaning-based part stands down, and the results look the same either way — so when a search turns up little, try the words the entry itself would use, such as a site name or an asset number.
The assistant searches your entries the same way when you ask it a question. If meaning-based matching is unavailable, it treats an empty answer as inconclusive rather than as proof that no entry fits. To learn how it draws on your entries, see the assistant.