Adding sources
Upload files or import web pages, notes and documents from Google Drive, Microsoft 365 and Outlook. Processing runs in the background.
Sources are the documents a brain answers from. Each source is converted to text, given a title and summary, and used in answers as soon as it shows Ready.
Upload files
Open Add sources
Open the brain, choose the Sources tab and click Add sources.

Choose files
Drag up to 10 files onto "Drop files here or choose them", or click it.
Wait for processing
Each file shows its state, such as Uploading… or Processing. Once everything is uploaded, you can leave the page.
A file with the same content is skipped (Already in this brain, skipped). A file with the same name and new content replaces the older version (Keeping a brain up to date).
Formats and limits
| Kind | Formats |
|---|---|
| Documents | PDF, Word (.docx), PowerPoint (.pptx), .odt, RTF, e-mails (.eml) |
| Spreadsheets | Excel (.xlsx, .xlsm, .xls), .ods, CSV |
| Text | text, Markdown, HTML, JSON, XML and other text formats |
| Images | PNG, JPEG, WebP, GIF, only in brains set to Open |
Among the formats that cannot be read are .doc, .ppt, .odp, .msg, Pages, Numbers and HEIC. Save such files as .docx, .pptx, .pdf or .png.
Other sources
- Web page: enter the URL under Address and click Import. Only publicly reachable pages; pages behind a sign-in or built purely with JavaScript return no text.
- Paste text: enter a Title (optional) and the Text, then click Add. Good for notes and short policies.
- Google Drive, Microsoft 365, Outlook (mail & calendar): appear when your workspace offers the connector; without a connection the tile shows Connect. From Outlook you import the text of an e-mail or one of its attachments (Connectors overview).
From the cloud, the text arrives as a snapshot, through your personal connection. Everyone who may read the brain sees this text, regardless of the permissions in the cloud storage. Custodos keeps documents from Google Drive and Microsoft 365 current, e-mails not (Keeping a brain up to date).
From a chat, you save a text draft to a brain as a copy with Save to Brain (The document beside the chat).
Images and scanned PDFs
Photos, screenshots and PDF pages without text are read by a model with image recognition, and the text is stored, not the image. This works only for direct uploads into a brain set to Open, and only if an admin has allowed a model that reads images. It costs extra credits. A chart on a page that also has text is not read.
Messages
| Message | What you can do |
|---|---|
| File too large (max. 20 MB). | Split or shrink the file. |
| File type is not supported. | Save it as .docx, .pdf or .png. |
| The file contains no readable text. | For a scan over 12 MB: shrink or split it. |
| This brain is set to "Strict" and the file contains protected data. | Remove the identifier or choose another brain. |
| Images only go into brains with "Open" data protection, … | Put the image into an open brain. |
| No allowed model can read images. … | Ask an admin to allow such a model, then Process again. |
| Your personal monthly budget is used up. … | Ask an admin for more budget, then Process again. |
| Your group keeps your data in Switzerland, … | Your group (on request) allows no processing outside Switzerland. You can still read the brain; Add sources is missing. Ask an admin. |
Process again is in the Actions menu of the document row. If a document shows Error, hovering over the status shows the reason.
Frequently asked questions
Usually the page is not public, requires a sign-in or builds its content with JavaScript. Save the page as a PDF and upload the file.
Only for Word, Excel and PowerPoint files up to 4 MB that you upload directly. They can be downloaded and edited in chat. For everything else the brain stores the text.