Creating a Knowledge Collection

Set up a new knowledge collection to organize your documents.

Overview

A collection is a named bucket of related documents that your workflows can search. Every document you upload lives inside exactly one collection, and every Knowledge node queries one collection at a time, so the collections you create define the units of search available across your workspace.

Creating a collection is a one-time setup step. Once it exists, you upload documents into it, and any workflow in the same workspace can query or embed into it through a Knowledge node.

Before You Start

You need access to a workspace where you have permission to manage knowledge.
Decide on the scope of the collection up front: one topic, one source system, or one document type.
Check your remaining storage in Settings -> Usage if you plan to upload large files immediately after creating the collection.

Steps

Open Knowledge from the sidebar. Spojit shows the list of collections in your current workspace.
Click + New Collection.
Enter a name (for example product-catalog or company-policies). The name is what you will select when configuring a Knowledge node, so keep it short and descriptive.
Click Create. The empty collection appears at the top of the list and is ready to receive documents.
Open the new collection and upload your first documents, or wire a workflow to embed into it.

Persistent vs Transient Collections

Spojit offers two kinds of collections:

Persistent collections are what you create here - named, long-lived buckets that any workflow in the workspace can embed into or query. Use them when the same documents need to be searchable across multiple workflows or over time (product catalogs, policy libraries, FAQ knowledge bases).
Transient collections are single-run only. You don't create them up front - any Knowledge node can select Transient from the collection dropdown to create one on the fly. Embed and Query nodes in the same workflow run share the transient collection, and Spojit cleans it up automatically when the workflow completes. Use this for one-off document processing like extracting structured data from an emailed PDF.

The rest of this article covers persistent collections. For transient usage see Using Knowledge Nodes.

Naming Conventions

Use lowercase, hyphenated names so they read clearly in node configuration menus.
Match the collection name to its content, not its purpose. contracts-2025 ages better than legal-lookup.
Avoid environment prefixes like prod- or staging-. Knowledge collections are workspace-scoped, so the workspace already conveys the environment.

Tips

Create separate collections for different content types. A focused collection produces more relevant search results than a single catch-all.
Plan a small number of collections at first. You can always split a collection later by re-uploading subsets into new ones.
Pick the collection name with the workflow author in mind, since they will see it in a dropdown when adding a Knowledge node.

Common Pitfalls

Creating one collection per file. Collections are meant to group related documents so queries can find the best match across them.
Renaming a collection that workflows already reference. Update the node configuration in each affected workflow, or queries will break.
Mixing internal and customer-facing content in the same collection. A query meant for support replies could surface confidential material.

Learn More

Knowledge base documentation