Uploading Documents to a Collection

Add documents to your knowledge collection for AI-powered search.

Overview

Uploading is how content enters the Knowledge Base. Files added to a collection are parsed, split into chunks, and converted into vector embeddings so that workflows can search them with natural language. Until a document has finished processing, it does not appear in query results.

You can upload directly through the Knowledge UI for one-off ingestion, or have a workflow add documents programmatically using a Knowledge node in embed mode. The same collection accepts both sources, so a workspace can mix manually uploaded reference material with documents that arrive through automation.

Before You Start

You need an existing collection. See Creating a Knowledge Collection.
Confirm the file is in a supported format (see the list below).
Check your remaining storage in Settings -> Usage before bulk uploads.

Supported Formats

Documents - PDF, DOCX, DOC, RTF, ODT, TXT, MD
Spreadsheets - XLSX, XLS, CSV, TSV, ODS
Presentations - PPTX, PPT, ODP
Email - EML, MSG
Web - HTML, XML
Images - PNG, JPG, JPEG, TIFF, BMP
Other - JSON, EPUB, RST

Uploading From the UI

Open Knowledge from the sidebar and click the collection you want to add to.
Click Upload, or drag and drop one or more files into the document list.
Wait for each file to finish processing. The status changes once embeddings are ready and the document becomes queryable.

Uploading From a Workflow

To ingest documents automatically (for example, every invoice that arrives by email), add a Knowledge node in embed mode and point it at the target collection. Pass the file content from an earlier step using a variable like {{ trigger.attachment }}. The workflow then handles parsing and indexing on every run.

Tips

Prefer text-based PDFs over scanned images where you can. Native text extracts cleanly and indexes faster.
For spreadsheets, give columns clear headers. Embeddings work better when each row reads as a coherent record.
When automating uploads from a workflow, name the document with something searchable (the invoice number, the contract ID) so it is easy to find later.

Common Pitfalls

Querying immediately after upload. Wait for processing to finish, otherwise the new document will not appear in results.
Re-uploading an edited file without removing the old version. Both will remain in the collection and compete in search results.
Uploading huge image-only PDFs. They take longer to process and consume more of your storage allowance than the text would suggest.
Uploading sensitive content into a collection that any workflow in the workspace can query. Use a dedicated collection for restricted material.

Learn More

Knowledge base documentation