Skip to main content
Home > Admin Menu > Settings > Expert Settings
Configure advanced search, chunk and text recognition settings for your VARIOS AI instance.
The Expert Settings tab is only displayed if you have enabled the Expert View in your account settings.

​
Hybrid Search Weighting

Configure the weighting between vector search and full-text search for the hybrid search mode. The sum of both weights should equal 100%.
An even distribution (50/50) is a good starting point. Increase the vector weighting for semantically demanding searches or the full-text weighting for exact keyword matches.

​
Chunk Configuration

Configure how documents are split into sections (chunks) for processing by the embedding model.
Changes to the chunk configuration may require existing documents in the knowledge base to be re-embedded. Depending on the amount of data, this can take time and incur costs.

​
Tika OCR strategy

Under OCR strategy, define how Apache Tika extracts text from PDF documents, including scanned ones. The setting applies to every text extraction in VARIOS AI: when embedding documents in knowledge bases, when reading files in the chat, during the DLP check of uploaded files, and for files read by the Microsoft 365 connector.
Higher OCR usage increases processing time per document. With OCR and text and OCR only, every page is rendered. For scanned PDFs with large pages, this can take many seconds per page.
A changed strategy applies to documents read in afterwards. Documents that are already embedded keep their text until they are embedded again. Files in the chat whose text has already been extracted keep it as well.