Skip to content
Skip to case study
Back to work
CASE STUDY / 03LinkedInX

Learning systems · 2026

Aletheia Learning

A study system that turns dense sources into connectable, reviewable, usable knowledge.

TL;DR

A parser that walks an Obsidian vault, extracts wiki-link connections and question/answer pairs marked in the text, and turns them into flashcards ready to review without leaving the notes system.

2 patternsWiki-links and Q::A cards detected
0 dependenciesCloud — runs on the local vault
*.mdInput and output format, no data lock-in
  • Concept extraction
  • Assisted flashcards
  • Native vault export
PythonMarkdownObsidian

01 / CONTEXT

The problem

Reading a dense source (a course, a paper, a technical book) produces loose notes that are rarely reviewed or connected to each other.

02 / DIRECTION

The solution

A parser that walks an Obsidian vault, extracts wiki-link connections and question/answer pairs marked in the text, and turns them into flashcards ready to review without leaving the notes system.

03 / EXECUTION

The process

01Architecture
02Prototype
03Iteration
04Delivery
  1. 01Defining a writing convention (Q::A, [[links]]) that would not break natural note-taking.
  2. 02Parser prototype with regular expressions over a real test vault.
  3. 03Iterating to tolerate malformed notes without silently losing content.
  4. 04Native export back into the vault so knowledge does not live in a separate tool.

04 / RESULT

Impact signals

2 patternsWiki-links and Q::A cards detected
0 dependenciesCloud — runs on the local vault
*.mdInput and output format, no data lock-in

05 / IMPLEMENTATION

A decision in code

obsidian_parser.py

import re
from pathlib import Path

def parse_markdown_vault(vault_path):
    """Scan obsidian markdown vault to extract connected nodes and flashcards."""
    vault = Path(vault_path)
    notes_processed = 0
    cards_extracted = []
    
    # Regex to match [[Wiki Link]] connections and Q::A flashcards
    link_pattern = re.compile(r'\[\[([^\]]+)\]\]')
    card_pattern = re.compile(r'^([^\n]+)::([^\n]+)')
    
    for note in vault.glob("**/*.md"):
        content = note.read_text(encoding="utf-8")
        links = link_pattern.findall(content)
        for line in content.splitlines():
            match = card_pattern.match(line)
            if match:
                cards_extracted.append({
                    "source": note.name,
                    "front": match.group(1).strip(),
                    "back": match.group(2).strip()
                })
        notes_processed += 1
    return {"vault_notes": notes_processed, "extracted_cards": len(cards_extracted)}

06 / LEARNING

Reflection

The clearest lesson was design, not code: any study system that does not respect the format you already think in is abandoned within two weeks.