Markdown Parser

Refers to text blocks through multiple levels of abstraction, starting at headers, all the way down to lines and individual character offsets. Also includes a hierarchical text segmentation algorithm to fit a token constraint, largest headers prioritized first. Has utility in RAG, embeddings, and MPC.

See demo for usage.

The MarkdownIndexer class will construct an explicitly-sectioned outline of the text through the headers.
A line starting with the least number of # characters will set the top level section, sections with more #'s will receive an extra '.'

1 : Global Convergence of Online Limited Memory BFGS
         1.0.0.1 : Abstract
   1.1 : Introduction
   1.2 : Algorithm Definition
      1.2.1 : LBFGS: Limited Memory BFGS
      1.2.2 : Online (Stochastic) Limited Memory BFGS
   1.3 : Convergence Analysis
   1.4 : Search Engine Advertising
      1.4.1 : Feature Vectors
      1.4.2 : Logistic Regression of Click-Through Rate
      1.4.3 : Numerical Results
   1.5 : Conclusions
   1.6 : Acknowledgments
   1.7 : Appendix A. Proof of Proposition 1
   1.8 : Appendix B. Proof of Lemma 2
   1.9 : Appendix C. Proof of Lemma 3
   1.10 : Appendix D. Proof of Lemma 4
   1.11 : Appendix E. Proof of Lemma 5
   1.12 : Appendix F. Proof of Theorem 6
   1.13 : Appendix G. Proof of Theorem 7
   1.14 : References
Markdown is 1020 lines long, with total character length 137372.