updated README

README.md

@@ -10,7 +10,7 @@ short_description: Chunker simple from Sergii Makarevych's SDDW-based Chunker
 
 # Chunker
 
-[Chunker] [https://huggingface.co/spaces/prashantmatlani/chunker], a supposed substitute for RAG, is a low-level "browse a file and extract" tool of knowledge extraction via Progressive Disclosure from a large and loaded source such as a book (articles, papers work too) - taking a cue from Sergii Makarevych's [work] [https://www.linkedin.com/pulse/new-way-encode-documents-ai-agents-navigable-trees-sergii-makarevych-a6cof/], in turn derived from Andrej Karpathy's [LLM Wiki pattern] [https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f]
+[Chunker] (https://huggingface.co/spaces/prashantmatlani/chunker), a supposed substitute for RAG, is a low-level "browse a file and extract" tool of knowledge extraction via Progressive Disclosure from a large and loaded source such as a book (articles, papers work too) - taking a cue from Sergii Makarevych's [work] (https://www.linkedin.com/pulse/new-way-encode-documents-ai-agents-navigable-trees-sergii-makarevych-a6cof/), in turn derived from Andrej Karpathy's [LLM Wiki pattern] (https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f)
 
 Chunker fulfills the base requirement of getting something along the lines of chunks and summaries as elaborated paragraphs from knowledge-source such as books. - though scaled up a little to include expanded Summary at Level 3 - worth hundreds of pages; called, although, "Jungian Chunker" it's usable for any pdf (though yet to be checked the max number of pages it can cover upto, but it probably runs into a few hundreds, and on the downside, its time-consuming)
 
@@ -26,14 +26,14 @@ The combined script - with two Phases, I (Extract and rewrite chunks - The "Leav
 * The "Children" Key: In the final JSON, each summary block now lists which leaf chunks belong to it; making it a Navigable Tree
 
 
-** The Logic of the Knowledge-Pyramid: **  
+**The Logic of the Knowledge-Pyramid:**  
 
 * L0 (Leaves): 1-2 pages of raw text rewritten
 * L1 (Clusters/Branches): Summary of 5 Leaves (~10 pages)
 * L2 (Chapters): Summary of 5 L1 Clusters/Branches (~50 pages)
 * L3 (Volume): Summary of all L2 Nodes (The entire book)
 
-** Phase I - Intelligent Chunking/Extract and rewrite chunks (The "Leaves") **
+**Phase I - Intelligent Chunking/Extract and rewrite chunks (The "Leaves")**
 
 > The Semantic Split: Instead of splitting - at fixed token boundaries, a cursor advances through the text and the LLM finds semantically complete split points, such as - at exactly 1000 characters, we give the LLM a 6000-character window and ask it to find the natural "Topic End" - locations where a topic is probable to end with another to probably begin - via "break_text".