Settings

Configure your preferences for the RAG system.

LLM Configuration

Loading LLM settings...

Document Embedding

Choose the way excel files are loaded and what document transformation strategy is used for vector db.
Text: Transforms everything as text, will be chunked after and we may lose context of headers. Fastest upload
Column: Each column is treated as separate document. Average upload speed
Row: Each row is treated as separate document. Average upload speed
Cell: Each cell is treated as separate document. Very slow upload

Document Retrieval

The regular expression used to detect and normalize codes (e.g., error codes) in the user query for re-ranking.
Configure the MMR retriever to balance relevance and diversity of search results.
The number of documents to fetch from the vector store to pass to MMR (1-100).
The number of documents to return after MMR diversification (1-20).
0.4
Controls diversity. 0 is most diverse, 1 is least diverse.
Case-insensitive code normalization
When enabled, code patterns (e.g., FP-04) are detected regardless of case.
Force lowercase queries (robust retrieval)
When enabled, queries are lowercased before retrieval, reducing accidental code-like matches (e.g., "AI 300").
Enable Advanced Re-ranking
Improves accuracy by using a more advanced model to re-rank search results. This may slightly increase response time.

Debugging

Enable Debug Mode
When enabled, debug buttons will be visible on the Assistant and Knowledge Base pages for troubleshooting and detailed analysis.