By default Edna answers questions but does not touch anything. This article covers turning on the ability to propose changes, and using a document as the starting point for one. Nothing is created without you approving it first, which is how the assistant is meant to work.
Read Edna, the AI assistant: what it can do first if you have not.
The "Edna can make changes" switch
This is a lab-wide setting, and only a PI can change it.
- Go to Settings in the sidebar.
- Open the Lab tab.
- Scroll to the Edna section.
- Turn on Edna can make changes.
You get a confirmation that Edna can now make changes. Turning it off gives the matching confirmation.
If you are not a PI, the setting is visible but not editable, and it says only a PI can change it.
What the switch does and does not do
The app is precise about this, and it is worth quoting the shape of it:
- Asking questions is always available on eligible plans, switch or no switch.
- Changes additionally require this switch.
- Every change still needs your confirmation in chat.
So the switch is not "let Edna act on its own". There is no setting anywhere that does that. The switch decides whether Edna may offer to change something. Whether it happens is still your call, one card at a time.
Turning it on is a low-risk decision for that reason. The thing it unlocks is Edna preparing a filled-in form for you to check, which is faster and less error-prone than typing the same thing yourself.
Attaching a PDF or photo in chat
You can attach a document to a chat message and have Edna read it into a proposal. This is the quick path for a single arriving delivery, as opposed to loading a whole existing inventory.
- Open Edna.
- Use Attach a PDF to import.
- Pick your file.
- Wait while it reads. It tells you it is reading the PDF.
- Review what it found, and approve.
Supported files
PDF, JPEG, PNG, and HEIC. Anything else is refused with a message telling you to attach a PDF or a photo in one of those formats.
What you will see
Edna extracts the inventory items it can find and presents them for review before anything is created. On each item you can set:
- Whether it is a reagent or a supply.
- The location, with a searchable picker.
- The vendor, choosing an existing one or typing a new one.
- No lot yet (add later) if the lot number is not on the document. See Lot numbers, CAS, hazard and expiration.
Nothing is created until you approve. The same propose-then-confirm rule applies here as everywhere else.
When it does not work
The messages are direct rather than vague, so you can tell what went wrong:
- "I could not read that PDF" means the file itself was a problem. Check it and try again.
- "No inventory items were found in that PDF" means it read fine but had nothing recognisable in it.
- If adding fails, it tells you nothing was created and you can try again. A failed attempt does not leave half your items in place.
Chat attachment or the guided import?
Both read documents, and they are for different jobs. The guided import is the one you will reach for most.
| Use | When |
|---|---|
| Guided import | The normal case. A delivery being unpacked, a stack of packing slips, a spreadsheet, a shelf you are finally recording. Many files at once, with location matching and bulk editing. |
| Attach in chat | The exception. A single document with a handful of items, when you are already in a conversation and it is quicker to ask. |
The guided import handles CSV spreadsheets, deduplicates locations across files, and lets you edit in bulk, none of which the chat path does. Reach for it for anything sizeable. Open Add Inventory in the top bar and choose Import inventory. See Import files: CSV, vendor PDFs and shelf photos.
It is not only for new labs. Its name comes from the Dashboard card new labs see, but the same flow is how inventory arrives from then on: every order, every delivery, every batch you never got round to entering. Edna's welcome message points new labs at it, which is why it can read as a one-time setup step. It is not.