Class: Ask::RAG::TextSplitter::Markdown

Inherits:
Base
  • Object
show all
Defined in:
lib/ask/rag/text_splitter/markdown.rb

Overview

Splits Markdown text by headers, preserving header hierarchy in metadata.

Each chunk starts at a header and includes all content until the next header at the same or higher level. Header context is tracked in metadata so retrieval results know which section they came from.

Examples:

splitter = Ask::RAG::TextSplitter::Markdown.new
chunks = splitter.split_documents(docs)
chunks[0].metadata[:headers] # => { h1: "Introduction", h2: "Background" }

Constant Summary collapse

HEADER_PATTERN = 
/^(#{'#'}{1,6})\s+(.+)$/

Instance Method Summary collapse

Methods inherited from Base

#initialize, #split_documents

Constructor Details

This class inherits a constructor from Ask::RAG::TextSplitter::Base

Instance Method Details

#split_text(text) ⇒ Object 



20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
 
# File 'lib/ask/rag/text_splitter/markdown.rb', line 20

def split_text(text)
  chunks = []
  current_header = {}
  current_lines = []

  text.each_line do |line|
    if (match = line.match(HEADER_PATTERN))
      # Save previous section
      if current_lines.any?
        chunks << build_chunk(current_lines.join, current_header)
      end

      level = match[1].length
      heading_text = match[2].strip

      # Clear lower-level headers
      current_header.select! { |k, _| HEADER_LEVELS[k] && HEADER_LEVELS[k] < level }
      current_header["h#{level}".to_sym] = heading_text

      current_lines = []
    else
      current_lines << line
    end
  end

  # Last section
  chunks << build_chunk(current_lines.join, current_header) if current_lines.any?
  chunks
end