> ## Documentation Index
> Fetch the complete documentation index at: https://docs.spojit.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Knowledge

> Manage your document collections for AI-powered search and extraction.

The **Knowledge** section is where you manage your workspace's document collections (the vector stores that power the [Knowledge node](/workflow-editor/nodes/knowledge) in your workflows). You can create collections, upload documents, view their status, and remove documents you no longer need.

## Collections

A **collection** is a named group of embedded documents that any workflow in your workspace can query. Collections are workspace-scoped, meaning all team members and workflows share the same collections.

### Creating a collection

<Steps>
  <Step title="Open the Knowledge page">
    Navigate to **Knowledge** in the sidebar.
  </Step>

  <Step title="Create a new collection">
    Click **New Collection**, enter a name and optional description, then click **Create**.
  </Step>
</Steps>

Collection names should be descriptive, for example `invoices`, `company-policies`, or `product-catalog`.

Each collection uses an [embedding model](/platform/models#embedding-models) to convert documents into vectors. The default is good for most cases; pick a different one only if you need higher quality or lower cost for high-volume ingestion. Once a collection is created, its embedding model is fixed.

### Viewing a collection

Click on a collection to see its documents. The detail page shows:

* **Document count**: how many documents are in the collection
* **Total size**: the combined size of all uploaded documents
* **Document table**: each document with its file name, type, size, chunk count, status, and upload date

You can filter documents by name or type, and sort by any column.

## Uploading documents

<Steps>
  <Step title="Open the collection">
    Navigate to **Knowledge** and click the collection you want to add documents to.
  </Step>

  <Step title="Click Upload Document">
    Click the **Upload Document** button in the top right.
  </Step>

  <Step title="Select a file">
    Drag and drop a file or click to browse. Supported formats include **PDF**, **Word** (docx, doc), **Excel** (xlsx, xls), **PowerPoint** (pptx, ppt), **CSV**, **TSV**, **JSON**, **XML**, **HTML**, **TXT**, **Markdown**, **RTF**, **Email** (eml, msg), **EPUB**, and **images** (png, jpg, tiff, bmp) with OCR.
  </Step>

  <Step title="Upload & Embed">
    Click **Upload & Embed**. The document is parsed, split into chunks, and embedded into the vector store. This may take a few seconds depending on file size.
  </Step>
</Steps>

### Duplicate file handling

If you upload a file with the same name as an existing document in the collection, you'll be prompted to:

* **Overwrite**: delete the existing document and replace it with the new one
* **Rename**: upload with a different file name
* **Cancel**: abort the upload

### Document status

| Status         | Meaning                                         |
| -------------- | ----------------------------------------------- |
| **READY**      | Document is embedded and available for queries  |
| **PROCESSING** | Document is currently being parsed and embedded |

## Deleting documents

To delete a document, click the trash icon next to it in the document table. This permanently removes the document and all its embedded chunks from the vector store.

<Warning>
  Deleting a document cannot be undone. The embedded chunks are permanently removed from the vector store.
</Warning>

## Deleting collections

To delete an entire collection, click the delete button on the collection card in the Knowledge overview page. This removes all documents and their embeddings.

## Workflow integration

Documents uploaded through the Knowledge page are available to any workflow using a [Knowledge node](/workflow-editor/nodes/knowledge) in query mode. Workflows can also embed documents programmatically using embed mode, and these appear in the collection alongside manually uploaded documents.

<Tip>
  For one-off document processing (like invoice extraction), use **Transient** collections in your workflow instead of a persistent collection. Transient collections are automatically cleaned up after the workflow completes. See [Knowledge node: Transient collections](/workflow-editor/nodes/knowledge#transient-collections) for details.
</Tip>