Knowledge Sources

Knowledge sources let you give assistants access to your own data. By connecting documents, files, or websites, you enable the AI to search and reference your information when answering questions — producing more accurate, context-aware responses grounded in your content.

Types of Knowledge Sources

nunq supports several types of knowledge sources:

File Uploads

Upload documents directly to nunq for processing and indexing. Supported formats include:

• PDF documents
• Microsoft Word documents (.docx)
• Microsoft Excel spreadsheets (.xlsx)
• Plain text files (.txt)

Uploaded files are parsed, split into chunks, and indexed so the AI can retrieve relevant passages during a conversation.

OneDrive Files and Folders

Connect your Microsoft OneDrive account to make files and folders available as knowledge sources. nunq indexes the selected files so the AI can search their content without requiring manual uploads.

SharePoint Sites and Pages

Link SharePoint sites or individual pages as knowledge sources. This is useful for organizations that maintain internal documentation, policies, or knowledge bases in SharePoint.

Websites

Provide a URL and nunq will crawl and index the website content. You can configure:

• Crawl depth — How many levels of links to follow from the starting page.
• Page limits — The maximum number of pages to index.

This is useful for making external documentation, help centers, or public knowledge bases available to your assistants.

Plain Text Input

Enter text directly as a knowledge source. This is a simple option for adding short reference material, guidelines, or context that does not exist as a separate file.

Attaching Knowledge Sources to Assistants

Once a knowledge source is created, you can attach it to one or more assistants. When an assistant has knowledge sources attached, the AI automatically searches them for relevant information before generating a response. See Assistants for details on configuring assistants.

Processing Status

After you add a knowledge source, it goes through a processing pipeline:

1. Pending — The knowledge source has been submitted and is queued for processing.
2. Indexing — The content is being parsed, chunked, and indexed for retrieval.
3. Completed — Processing is finished and the knowledge source is ready for use.

You can monitor the processing status from the Knowledge Sources section.

How RAG Works

nunq uses retrieval-augmented generation (RAG) to incorporate your knowledge sources into AI responses. When you ask a question:

1. The AI searches your attached knowledge sources for passages relevant to your query.
2. The most relevant passages are included as context alongside your message.
3. The AI generates a response that draws on both its general knowledge and the retrieved content.

This approach allows the AI to provide answers that are grounded in your specific data, reducing hallucinations and improving accuracy for domain-specific questions.

OCR Support for Scanned Documents

Knowledge sources that contain scanned documents or images of text are processed using optical character recognition (OCR) to extract readable text. nunq supports two OCR modes:

• Default OCR — Built-in text extraction suitable for most documents.
• Azure Document Intelligence — A more advanced OCR option that provides higher accuracy for complex layouts, tables, and handwritten content. This can be enabled in the knowledge source configuration if your organization has it configured.