Using Space Brain
The Library: files and folders
Upload anything once, browse it like folders, and let Space work out how each file can be used.
The Library is the Files area of Brain. Upload your documents and Space does the rest. You never need to decide whether something is “a knowledge file” or “a data file”.
Uploading
- Open Brain.
- Choose Upload, then pick files, or a whole folder.
- Watch the status of each file update.
Folder structure is kept: upload Finance/2026/revenue.csv and you get the same folders. Files are processed in the background, one at a time, so one failure never blocks the rest of a big upload.
What the status tells you
| Label | Meaning |
|---|---|
| Processing | Being read. It will update by itself. |
| Searchable document | Its text can be searched, and answers can cite it. |
| Table data | It has rows and columns agents can filter, sort and total. |
| Graph | It also has entities and facts in the knowledge graph (if that's on). |
| Needs attention | Something failed. Open the file's details to see why and Retry. |
Your original file stays downloadable at every stage, even if processing fails.
How Space handles different files
| What you upload | What happens |
|---|---|
| A narrative document: policy, manual, report, contract | The text is extracted and indexed, so a question finds the right passage. Scanned pages are read with text recognition, if enabled. |
| A table of records: a CSV, or a spreadsheet of rows | The rows and columns are kept exactly. They're analysed with calculations, never guessed at. |
| A mixed document: a report with narrative and tables | Both: the narrative is searchable, and the tables are kept whole and readable. |
| A PDF or scan with a table in it | The table is read and kept, and marked as extracted from a PDF or image, so you know to check important numbers against the original. |
| Something unsupported or unreadable | Kept as is, with a clear note of what couldn't be done. |
Supported formats include PDF, Word, Excel, CSV, plain text, HTML and Markdown, plus images (with text recognition).
Correcting how a file is treated
Space guesses from the file's content, not just its extension, and the guess is usually right. If it isn't, open the file's details and choose a different processing mode: Auto, Document or Table. Reprocessing clears old results so nothing stale remains.
File details
Click a file to see:
- Its capabilities and a preview.
- The processing mode and a Retry button.
- A Share control (see below).
- For tables from PDFs or images, a reminder to check the original.
Sharing a file
New uploads are yours. To give someone access, open the file and share it with a person or a group. They get read-only access. Adding a file to a knowledge base does not share it.
Moving, deleting, duplicates
- Moving files or folders keeps their identity, so citations and knowledge bases keep working, and nothing is re-read.
- Deleting a file removes everything derived from it: its search entries, summaries, graph facts, ratings and the answers that cited it.
- Duplicates with identical content are treated as one in search results, which names the other copies.
Limits worth knowing
- Your administrator sets a maximum upload size.
- Very large spreadsheets (around a million cells or more) stay downloadable but aren't analysed as tables.
- Tables inside born-digital PDFs are best when drawn with ruled lines. Borderless tables stay in the text unless your organization uses an advanced document reader.
Paths agents see
Your personal files appear to agents as /me/…. Files shared with you appear under /library/…. You'll see these paths if you ask an agent to read or save a file.