Logo IconGuided Mind
v2.4Sign In
RAG Wizard

Step 2 — Knowledge Sources

Upload documents and configure data sources for your RAG system.

Add your knowledge base content through document uploads, URL sources, or existing document libraries.

Adding Documents

Knowledge sources are the foundation of your RAG system. The quality and relevance of your search results depend directly on the documents you provide. You can add content through direct uploads or by selecting from previously stored documents.

You can add content to your RAG project in two ways:

SourceDescription
Document UploadUpload files directly (PDF, HTML, CSV, DOCX, MD, XLS)
Document LibrarySelect from previously uploaded documents stored in your account

Supported File Formats

Different file formats require different processing pipelines. Understanding which format you're working with helps set expectations for text extraction quality and processing time.

FormatExtensionNotes
PDF.pdfText extraction with optional OCR
HTML.html, .htmScript and style tags stripped
CSV.csvTabular data converted to text
DOCX.docxWord documents with formatting
Markdown.md, .mdxPreserves heading structure
Excel.xls, .xlsxSpreadsheet data as text

Document Library

The Document Library stores all previously uploaded documents in your account, making them available for reuse across multiple RAG projects. This eliminates the need to re-upload common reference materials and keeps your knowledge base consistent. Shared documents from team storage are also accessible through the library.

Previously uploaded documents appear in your Document Library and can be reused across projects. This avoids re-uploading the same files. Shared documents from team storage can also be accessed here.

Best Practices

Following these guidelines ensures optimal processing quality and retrieval performance:

  • Upload documents from a single domain for best retrieval accuracy
  • Keep individual files under 50 MB
  • Use Markdown when possible for better structure preservation
  • Review extracted text after upload to verify quality