legora

Generated using QuicKB, a tool developed by Adam Lucek.

QuicKB optimizes document retrieval by creating fine-tuned knowledge bases through an end-to-end pipeline that handles document chunking, training data generation, and embedding model optimization.

Chunking Configuration

• Chunker: RecursiveTokenChunker
• Parameters:
  • chunk_size: 400
  • chunk_overlap: 0
  • length_type: 'character'
  • separators: ['\n\n', '\n', '.', '?', '!', ' ', '']
  • keep_separator: True
  • is_separator_regex: False

Dataset Statistics

• Total chunks: 57,405
• Average chunk size: 52.6 words
• Source files: 604

Dataset Structure

This dataset contains the following fields:

• text: The content of each text chunk
• source: The source file path for the chunk
• id: Unique identifier for each chunk