How to parse and tally different header levels in LangGraph state data?

Other Questions langgraph
1

I’m building a LangGraph application for automated document creation. I need to extract and count various header types, including main sections, subsections, and sub-subsections, from a string stored in my state dictionary under the key “report_structure”.

state_data = {
    'report_structure': 'Final document structure:\n\n'
                       '**Chapter 1: Financial Planning Basics**\n\n'
                       '1.1 Understanding budgets and their role in planning\n'
                       '1.1.1 Key principles of effective budgeting\n'
                       '1.2 Overview of financial planning for professionals\n\n'
                       '**Chapter 2: Investment Strategies**\n\n'
                       '2.1 Risk assessment and portfolio management\n'
                       '2.1.1 Real-world examples of risk management\n'
                       '2.2 Retirement planning considerations\n'
                       '2.2.1 Statistical data on retirement savings\n'
                       '2.3 Emergency fund strategies\n'
                       '2.3.1 Common mistakes in emergency planning\n'
}

I’m struggling to differentiate between main chapters, such as “Chapter 1”, subsections like “1.1”, and sub-subsections like “1.1.1”, since they all follow similar numbering patterns. What would be the best method to count each header type separately?

2

Your structure looks good for regex parsing. Each hierarchy level has unique patterns you can match against.

For chapters, use \*\*Chapter \d+:.*?\*\* to catch the bold markdown. Subsections are just \d+\.\d+ and sub-subsections are \d+\.\d+\.\d+. I’ve dealt with similar parsing issues in technical docs - counting decimal points in the numbering scheme works every time.

Watch out for partial matches though. Use word boundaries and be specific about what comes before/after each pattern. I’d grab all matches with re.findall() first, then use len() to count them instead of trying to count directly with regex.