Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.iblueprint.ai/llms.txt

Use this file to discover all available pages before exploring further.

iBlueprint’s Knowledge Base feature lets you ingest documents, web pages, and cloud files, then query that content inside Blueprint nodes using keyword or semantic (RAG) search. When an AI node needs to answer questions grounded in your own data — a product manual, a support knowledge base, or a set of internal policies — attaching a knowledge base gives it accurate, up-to-date context rather than relying on the model’s training data alone.

​
Create a knowledge base

1

Open Knowledge Bases

From the main navigation, go to Knowledge Bases and click New knowledge base.
2

Name your knowledge base

Give it a descriptive name (for example, Product Documentation – v3). The name helps you identify it when attaching it to Blueprint nodes later.
3

Add content

Choose one or more of the ingestion methods below. You can mix methods within the same knowledge base.
4

Trigger embedding

After uploading, iBlueprint chunks the content and generates vector embeddings. You can monitor the embedding status from the knowledge base detail page. Semantic search is available once the status shows Complete.

​
Add content

​
Upload documents

Drag and drop files or use the file picker to upload:
  • PDF (.pdf)
  • Microsoft Word (.docx, .doc)
  • Plain text (.txt)
  • Markdown (.md)
iBlueprint automatically parses each file, splits it into chunks sized for your embedding model, and stores the chunks ready for retrieval.

​
Crawl a website

1

Click Add URL

In the knowledge base editor, click Add URL under the Web crawler section.
2

Enter the starting URL

Paste the URL you want to crawl (for example, https://docs.your-company.com).
3

Configure the crawl

Set the maximum depth and any URL path filters to control which pages get ingested. The crawler follows internal links starting from your URL.
4

Start the crawl

Click Crawl. iBlueprint fetches and indexes each page, updating progress in real time. You can re-crawl at any time to pick up new or changed pages.

​
Sync from cloud storage

iBlueprint can pull files directly from Google Drive or OneDrive.
  1. Click Add cloud source → Google Drive.
  2. Authorize iBlueprint to access your Drive.
  3. Select a folder or individual files to sync.
  4. iBlueprint downloads the files, converts them where necessary, and indexes the content.
Cloud sync creates a snapshot at the time of ingestion. Re-sync manually or set up a scheduled Blueprint trigger to keep the knowledge base current as your files change.

​
Use a knowledge base in a Blueprint

Once your knowledge base contains indexed content, attach it to a Blueprint node:
  1. Open a Blueprint and add or select a Knowledge base search node (or an AI node that supports RAG).
  2. In the node settings, choose your knowledge base from the Knowledge base dropdown.
  3. Select a search mode (see below).
  4. Connect the node’s output to whatever comes next in your workflow — typically an AI node that uses the retrieved chunks as context.

​
Search modes

Converts the query into a vector embedding and retrieves the chunks whose embeddings are closest in meaning, even if they use different words. Semantic search requires that the knowledge base has been fully embedded (status Complete) and an embedding-capable AI connection is configured.This mode is the best choice for natural-language questions where the exact phrasing in the documents is unknown.
For most customer-facing AI workflows, semantic search produces higher-quality answers. Use keyword search when you need deterministic, exact-match retrieval or when embedding costs are a concern.

​
Monitor embedding status

The knowledge base detail page shows the current embedding status for each document set:
StatusMeaning
noneContent added but not yet embedded
partialSome chunks have been embedded; others are pending
completeAll chunks are embedded and ready for semantic search
You can also see the total number of embedded chunks and the embedding model used.