Feeding your AI the right data is essential to generating relevant, high-quality responses. In this tutorial, you’ll learn the differences between direct file uploads, searchable knowledge bases, and dynamic vector stores in StackAI. Each option is designed for different document sizes, use cases, and performance needs—so you can build workflows that are both scalable and reliable.

Summary

• Files Node: upload full PDFs, Word docs, or text files for direct processing

• Adjust chunk size and overlap to optimize model input and control citations

• Enable text/image extraction and advanced parsing for charts and tables

• Use public file uploads (user input) or private file nodes (internal reference)

• Use URL scraping, audio, or image nodes for non-textual inputs

• Difference between Files Node vs. Knowledge Base:

  
  

  • Files: send entire doc to model

  • Knowledge Base: send only relevant excerpts via search

  
  

• Configure knowledge base indexing: top results, chunk size, page matching, etc.

• Expose knowledge base search to end users as part of your app interface

• Use dynamic vector stores for very large docs that can’t be indexed ahead of time

• Choose data strategy based on:

  
  

  • Document size

  • Context limitations

  • Risk of hallucination

  • Precision of query needs

  
  

• Next up: put it all together to build accurate, context-aware AI workflows