Skip to main content

​
Creating Knowledge Bases

iBlueprint.ai also lets you build knowledge bases—collections of documents and data sources that power retrieval‑augmented generation (RAG) for your prompts and chatbots. A knowledge base stores structured or unstructured content, chunked into passages and vectorised so that the AI can search and use it during responses.

​
Accessing the Knowledge Base Workspace

  1. Navigate to Org Knowledge – In the left sidebar, click Org Knowledge and then select the My Knowledge tab. This page lists your existing knowledge bases. Each entry shows whether it is vectorised and how many document sets it contains.
  2. Create a new knowledge base – At the bottom of the sidebar, click the New button and choose New Knowledge Base. iBlueprint creates an empty knowledge base and redirects you to its detail page. Alternatively, open an existing knowledge base by clicking its card.
Note: At the time of writing, the dedicated /knowledge-base/new page is still under development and may not display a form. If the new base page is blank, return to the My Knowledge tab and open an existing knowledge base to explore features.

​
Document Sets and Source Types

Inside a knowledge base, content is organised into document sets (folders). Each document set can ingest data from different sources. Click the ➕ icon next to Document Sets to choose among several source types:
Source TypeDescription
File Upload (Mixed)Upload PDFs, Word documents, spreadsheets, presentations, images, audio, video or plain‑text files. iBlueprint extracts text and splits it into chunks automatically. Images undergo OCR and audio/video are transcribed.
Manual EntryEnter plain text or markdown directly into the web editor. Useful for short snippets, notes or custom content. You can name the document and type or paste the contents.
JSONPaste or upload JSON data. iBlueprint parses the JSON to extract key/value pairs and stores them in structured form.
API EndpointProvide a RESTful API endpoint. iBlueprint fetches the data at crawl time, saving the responses for retrieval. You may need to specify parameters or authentication tokens.
GitHub/GitLab RepositoryConnect a public or private repository. iBlueprint indexes code files, README files and other documentation from the repo. This is ideal for technical knowledge bases.
SQLite DatabaseUpload a .sqlite file. Tables and rows are extracted and stored as structured entries.
Cloud Storage (Google Drive, OneDrive, SharePoint)Authorise access to your cloud drive and select documents or folders to index.
Spider WebsiteProvide a website URL. iBlueprint will crawl the site and index its pages for retrieval (respecting robots.txt rules).
Public DataChoose from curated public datasets offered by iBlueprint (e.g., news articles, scientific papers).
When creating a document set, you must give it a Name and optional Description. After selecting a source type, additional fields appear—for example, you can choose files to upload or supply a repository URL. Click Create to add the set. Each set appears in the Document Sets panel, where you can expand it to view individual documents.

​
Automatic Chunking and Vectorisation

Once documents are added to a set, iBlueprint processes them automatically:
  1. Chunking – Text is broken into manageable passages (chunks) based on structure (paragraphs, headings) and token length. This granularity ensures that retrieval queries return relevant snippets rather than entire files. For code or tables, the system preserves logical blocks.
  2. Vectorisation – Each chunk is encoded into a high‑dimensional vector using language‑model embeddings. These vectors enable efficient similarity search during retrieval. A note below the upload form explains that documents “will be automatically chunked and vectorised after creation for RAG search and structured retrieval.”
There is currently no user‑exposed setting to adjust chunk length or choose different embedding models; iBlueprint handles these details for you. After processing, the knowledge base card on the My Knowledge page will show a “Vectorised” status.

​
Managing and Searching the Knowledge Base

After processing, you can:
  • Preview documents – Click a document in the set to view its text or data in the main panel. This allows you to verify that the content was extracted correctly.
  • Run searches – Use the Search KB bar to query the knowledge base. The AI uses semantic search over the vectorised chunks to return the most relevant passages.
  • Integrate with prompts/chatbots – In the prompt editor or chatbot configuration, select your knowledge base as a data source. The AI will then reference the indexed content when generating responses.
By organising information into document sets and letting iBlueprint handle chunking and vectorisation, you can build robust knowledge bases that augment your prompts and chatbots with domain‑specific context.