Docs menu
Document collections
Unstructured files, indexed so that Chat can answer from them.
In short
A document collection is a folder of files, on Google Drive, Amazon S3, OneDrive or SharePoint, registered in Northstar. Its PDF files are indexed so that Chat can answer from their content and cite them. Collections are governed like datasets: teams get access to them, and Chat only searches the collections you are allowed to use.
Collections
To create a collection, pick a drive, browse its folders, and register a folder as a collection.
- A collection points to that folder. The files stay where they are.
- Collections are separate from datasets: they hold unstructured text, not tables.
- Access is granted per team, like any other resource. See teams and resources.
Sync and indexing
A sync reads the folder and indexes its files. Each file then has a status:
| Status | Meaning |
|---|---|
| Indexed | The text was extracted and Chat can search it. |
| Pending | The file is waiting to be indexed. |
| Failed | The file could not be indexed, for example a scanned PDF with no extractable text. |
| Skipped | The file is not a PDF. Only PDF files are indexed. |
| Deindexed | You removed the file from Chat. It stays out until you reindex it or the source file changes. |
- Resync now reads the folder again. Unchanged files are not indexed twice.
- You can reindex or deindex a single file, or deindex the whole collection.
- Delete collection removes the collection and its index from Northstar. The source files are never touched.
With Chat
- In Chat, mention a collection to ask questions about it. Answers cite the documents they come from, and say plainly when the documents do not contain the answer.
- On a collection's page, the chat bubble answers from that collection only.
- Chat only searches collections where you have the Ask AI right.
FAQ
Why is my Word or Excel file skipped? Only PDF files are indexed. Export the document to PDF, or register a tabular file as a dataset instead.
Why did a PDF fail? Most often it is a scan: the pages are images with no text to extract.
Does deleting a collection delete my files? No. It removes the collection and its index from Northstar; the files in your drive are not changed.