Documents

Uploading documents

Add PDF, DOCX and XLSX files to your library, the real size limit, and what happens while a document is processed.

Everything Spyral knows about a company starts with a document in the library. Uploading is the main way to get one there.

Where to upload

Open Library in the sidebar and choose Upload in the top right. The keyboard shortcut Ctrl+U takes you to the same page from anywhere in the app.

The page is titled Upload Documents.

What Spyral accepts

• PDF, DOCX and XLSX only.
• 10 MB maximum per file.

Spyral checks that the file really is what its name claims. Renaming a .doc to .docx will not get it past the check.

If you need to bring in a file larger than 10 MB, connect your cloud storage instead. Files imported that way are allowed to be much larger. See Importing from Google Drive, Dropbox and Gmail.

Uploading

1. Open the upload page.
2. Decide who can see the documents using the Document Access panel on the right. See below.
3. Optionally set the Company link for the whole batch. See below.
4. Drag your files onto the dropzone, or click it to browse. You can add several at once.
5. Each accepted file appears in a list with its size and a Pending label. Hover a pending file and use the X to remove it.
6. Choose Upload All.

Files are uploaded one after another rather than all at once, so a large batch takes a little longer than you might expect.

Choosing who can see the documents

The Document Access panel sets the visibility for everything in the batch.

• Global is the default. Everyone in your firm can see the document.
• Project limits it to the members of a project you pick. Choose All Members, or Specific and tick the people.

The Project button is greyed out when your firm has no projects yet, without saying why. Create a project first if you need project level access.

You can change access later from the library. See The document detail view.

Linking the batch to a company

The Company link panel applies to every file in the batch.

• Auto-detect is the default. Spyral works out the company from the contents of each document while it processes it.
• Existing company links every file in the batch to a company you already have.
• New company creates the company first, from a name and optionally an RCS number and share capital, then links the batch to it.

Choosing a company explicitly is worth doing when you already know it. The document appears on the company profile immediately instead of waiting for processing to finish, and your choice overrides whatever Spyral would have detected.

What the statuses mean

Each file moves through Pending, Uploading, Processing, then Complete or Error.

Processing is where the real work happens. Spyral reads the text, works out the document type and the company, and indexes the content so the assistant can answer questions about it. A document only becomes answerable once processing has finished.

A born digital PDF, a Word file or a spreadsheet is usually done in seconds. A scanned document takes longer, normally a minute or two.

You can leave the page at any point. Processing continues on the server, and the library updates itself when the document is ready. If you wait on the upload page for more than about two minutes, the file keeps its Processing label and a note appears saying it will appear in your library once it is ready. That is not an error.

Scanned and photographed documents

If a PDF has no text in it, which is normal for anything scanned or photographed, Spyral reads it with OCR automatically. You do not need to do anything or tick anything. It reads English, French, German, Dutch and Italian.

When a document sits on Processing

If a document has been processing for more than about fifteen minutes, Spyral notices and retries it by itself. You do not need to do anything.

After two automatic retries that do not finish, the document is marked Failed and you can retry it yourself with Rescan.

When a document fails

A failed document shows a Failed label in the library. Hover it to see the reason. The reasons and what they mean:

• Unsupported file type. The file is not a PDF, DOCX or XLSX.
• The file could not be downloaded from storage. A temporary fault. Rescan it.
• The pages could not be rendered for scanning. The PDF could not be turned into images to be read. It is usually damaged.
• No readable text was found. The scan may be blank, upside down or too poor to read.
• Text extraction took too long and was stopped. Usually a very long or very heavy scan. Split it and upload the parts.
• The document could not be read. The file may be corrupted or password protected. Remove the password and upload it again.
• Too little readable text was found in the document. A cover sheet or a page that is almost entirely an image lands here.
• The extracted text could not be prepared for search or The document could not be indexed for search. Rescan it.
• The scanning service is misconfigured. Contact support.
• Processing did not complete after several attempts. Rescan it.
• Processing failed. Rescan it, and contact support if it fails again.

Rescanning a document

Rescan is the circular arrow in the row's hover actions. It puts the document through the whole analysis again: reading the text, working out the type, detecting the company, and indexing.

Use it when a document failed, or when it finished but nothing was classified.

A rescan replaces the current classification with the new result. It is limited to three rescans per hour per person and ten per hour across the firm, so use it deliberately rather than on everything.

Spyral does not detect duplicates

Uploading the same file twice creates two separate documents. They are both classified, both indexed and both appear in search results and in the assistant's answers.

Delete the extra copy yourself if you upload something twice.

Other ways a document reaches the library

• Attaching a file inside a project chat and choosing to add it to the knowledge base puts it in the library as well.

• The annual accounts bulk upload, used when you first set a company up, is a separate route with its own rules and its own review queue. Files added there do not behave like library uploads.
• Connected cloud storage. See Importing from Google Drive, Dropbox and Gmail.