Mindmap HTML Generator

Source: https://github.com/aipoch/medical-research-skills

Mindmap HTML Generation

Overview

First organize input into a clear tree hierarchy, then call scripts/generate_mindmap_html.py to generate offline HTML. Prioritize reusing existing heading structure; do not modify visual code first — get structure correct, then let the script generate the output.

Tip: The generated HTML page now includes a top toolbar: Export PNG / Fullscreen. Content generation logic remains identical.

Workflow

1. Determine Input Source

• If the user has already provided Markdown headings or lists, render directly.
• If the user provides meeting notes, long text, or chat logs, first organize into Markdown hierarchy, then render.
• If the user provides a JSON tree, confirm fields are text / children, then render directly.

2. Organize Structure

• Prioritize preserving the user's original heading hierarchy.
• If the original text has no headings, first extract 1 main topic and several level-1 and level-2 branches.
• If the original is plain paragraphs or sentences, the script adaptively splits by blank lines, line breaks, colons, and periods.
• Recommended depth is 2-4 levels to avoid pages becoming hard to read.
• Keep node text as short phrases; do not stuff entire paragraphs into nodes.

Content Richness Requirements

• The default goal is not just "producing a map" but "complete information, clear hierarchy, usable for review, presentation, or reporting".
• Level-1 branches under root should typically be at least 6; if the topic is information-dense, expand to 8-12.
• Each level-1 branch should preferably have 3-6 level-2 nodes; do not force equal branch lengths for neatness.
• If a level-2 node contains methods, conditions, results, mechanisms, impacts, exceptions, or evidence, continue splitting to level-3 nodes.
• For information-dense paragraphs, prioritize splitting into more nodes rather than compressing into one summary sentence.
• For long texts, papers, and meeting notes, do not just keep section names — also include the actual information points under each section.
• Node text must be concise, but conciseness does not mean deleting details.

Information Extraction Priority

• Prioritize extracting: background, problem, objective, subject, method, materials, process, key variables, results, mechanism, conclusion, significance, limitations, applications.
• If the source text contains causal, conditional, parallel, or contrasting relationships, split into multiple nodes.
• If the source text contains proper nouns, core mechanisms, metrics, sample sizes, models, drugs, groups, or key conclusions, prioritize preserving them.
• If a level-1 node name is too abstract, add specific child nodes to carry the details.
• If content can be split into "subject / action / result / significance", prefer this structure.

Paper Input Enhancement Rules

• When processing papers, PDFs, or research abstracts, do not just extract section names — include information points within each section.
• Cover at minimum: research background, research question, experimental design, key results, mechanism chain, conclusions and implications.
• If paper information is sufficient, additionally include: research subject, sample size, model, groups, experimental methods, key molecules or pathways, intervention, efficacy, limitations.
• If PDF auto-extraction is too coarse, generate intermediate Markdown first, then restructure before rendering.

Prohibited Behaviors

• Do not distribute branches evenly just for neatness.
• Do not just restate headings without adding information beneath them.
• Do not write only one vague summary sentence for each of "Background, Methods, Results, Conclusion".
• Do not omit key numbers, key steps, key mechanisms, or key experimental subjects.
• Do not sacrifice information density for brevity.

3. Configure Script

• Edit the CONFIG block at the top of scripts/generate_mindmap_html.py.
• Common configuration options:
  • input_path: Input file path.
  • output_html_path: Output HTML path.
  • output_markdown_path: Intermediate Markdown path.
  • page_title: Page title.
  • fallback_title: Fallback title when no root heading exists.
  • auto_bold_levels: Heading levels to bold by default, recommend keeping [1, 2].
  • plain_text_root_max_children: Maximum level-1 branches when auto-splitting plain text.
  • plain_text_child_max_children: Maximum child items per level-1 branch when auto-splitting plain text.

4. Run Script

Run:

python scripts/generate_mindmap_html.py

The script will:

1. Read the input file.
2. Parse tree structure.
3. Generate intermediate Markdown.
4. Call npx markmap-cli to export offline HTML.
5. Inject fixed layout and color rules.

5. Validate Results

• Open the generated HTML, confirm root node, level-1 branch order, and text display are correct.
• Confirm only level-1 and level-2 headings remain bold; level-3 and deeper are not bold.
• Confirm branch counts and child counts follow content variation, not mechanical equal lengths.
• Self-check that level-1 branches sufficiently cover the topic.
• Self-check that key numbers, mechanisms, subjects, and conclusions are not omitted.
• Self-check for multiple level-1 branches with identical structure but empty content.
• If nodes are too long, go back to compress input text, then re-run the script.
• If npx or markmap-cli is missing, install the Node.js environment first.

Input Rules

• Supports .md, .txt, .json.
• Markdown: prioritizes ATX headings (# to ######), then indented lists.
• Plain text prioritizes splitting by blank lines, line breaks, colons, periods, and other punctuation into unequal-length branches and children.
• JSON must be a single tree with fields text and children.
• Output path, page title, and fallback title go in CONFIG — do not pile CLI arguments.