🌿 mindgrove

Analyze, classify, and auto-link any markdown vault. Role detection, sophistication scoring, and bidirectional similarity linking — for knowledge bases at scale.

Why this exists

When your markdown vault hits 500+ files, it becomes a graveyard. Notes pile up with no metadata, no connections, no way to know what's related to what.

mindgrove crawls your vault, analyzes every file, and does three things automatically:

1. Tags — injects rich YAML frontmatter: role, domain, sophistication score, capabilities, techniques
2. Scores — rates each note's complexity from standard → elite using a weighted scoring engine
3. Links — builds a similarity matrix across all files and writes bidirectional [[wikilinks]] between related notes

Built from a real workflow used to manage 1,300+ AI system prompts in Obsidian. Extracted into a universal library with zero vault-specific assumptions.

Works everywhere markdown lives

| Environment | YAML Frontmatter | Wikilinks | Visual Graph | |---|---|---|---| | Obsidian | ✅ | ✅ | ✅ Graph View | | Foam (VS Code) | ✅ | ✅ | ✅ Foam Graph | | Logseq | ✅ | ✅ | ✅ Graph View | | Dendron | ✅ | ✅ | ✅ Tree View | | Plain markdown folder | ✅ | ✅ | — (data is there, use any renderer) |

No matter where your notes live — mindgrove enriches them. The [[wikilinks]] it writes are understood by every major PKM tool, so your graph appears automatically next time you open the vault.

Install

npm install -g mindgrove

Or as a library in your project:

npm install mindgrove

CLI Usage

# Analyze any markdown folder
mindgrove analyze ./my-vault

# Works directly on your Obsidian vault
mindgrove analyze /path/to/obsidian-vault

# With custom options
mindgrove analyze ./my-vault --threshold 20 --max-connections 8

# Dry run — analyze only, no file writes
mindgrove analyze ./my-vault --dry-run

# Skip linking (just tag files, no wikilinks)
mindgrove analyze ./my-vault --no-linking

What gets written to your files

Each markdown file gets a YAML frontmatter block injected (or updated):

---
type: note
analyzed: 2026-04-18 21:00
role: prompt-engineer
domain: ai-ml
sophistication: advanced
size-category: large
char-count: 4821
token-estimate: 1206
capabilities:
  - reasoning
  - code-generation
  - optimization
techniques:
  - chain-of-thought
  - role-playing
  - constraints
tools:
  - web-search
  - code-execution
Connected Notes:
  - "[[RAG Specialist Agent]] (score: 22 | chain-of-thought, ai-ml)"
  - "[[Meta Orchestrator]] (score: 18 | role-playing)"
status: active
---

Open your vault in Obsidian after running mindgrove — every connected note pair appears as an edge in Graph View automatically.

Library Usage

import { analyzeVault } from 'mindgrove';

const result = await analyzeVault('./my-vault', {
  similarityThreshold: 15,
  maxConnectionsPerFile: 12,
  enableLinking: true,

  // Extend with your own patterns
  customRoles: {
    'finance-agent': /trading|portfolio|equity/i,
  },
  customDomains: {
    'fintech': /\bfinance\b|\btrading\b|\bstocks\b/i,
  },
});

console.log(`Processed: ${result.processed} files`);
console.log(`Connections: ${result.connections} links written`);

Configuration

| Option | Default | Description | |---|---|---| | similarityThreshold | 15 | Min score to create a link | | maxConnectionsPerFile | 12 | Cap links per file | | enableLinking | true | Run Phase 2 similarity linking | | excludeFolders | ['.obsidian', 'Templates', ...] | Folders to skip | | domainDetection | 'content' | 'content' or 'path' | | customRoles | {} | Add your own role patterns | | customDomains | {} | Add your own domain patterns | | customCapabilities | {} | Add your own capability patterns | | customTechniques | {} | Add your own technique patterns | | dryRun | false | Analyze without writing files | | reportFormat | 'markdown' | 'text', 'json', or 'markdown' |

Sophistication scoring

mindgrove scores every note from standard → elite based on a weighted formula:

score = (techniques × 2) + (constraints × 1.5) + (capabilities × 1) + (tools × 1)

| Score | Label | |---|---| | 25+ | elite | | 18–24 | advanced | | 12–17 | sophisticated | | 6–11 | intermediate | | 0–5 | standard |

Filter by sophistication: elite in Obsidian's search to instantly surface your best notes.

Real-world example

This library was extracted from a personal Obsidian vault containing 1,300+ AI system prompts. Running mindgrove analyze on that vault:

• Tagged all 1,300 files with role/domain/sophistication in ~4 minutes
• Found 847 high-quality similarity connections above threshold 15
• Wrote bidirectional [[wikilinks]] into every connected pair
• Enabled Obsidian Graph View filtering by sophistication: elite to instantly surface the best prompts

Project structure

mindgrove/
├── src/
│   ├── index.js            ← public API exports
│   ├── config.js           ← fully overridable DEFAULT_CONFIG
│   ├── analyzer.js         ← Phase 1 (tag) + Phase 2 (link) orchestrator
│   ├── similarity.js       ← weighted scoring matrix engine
│   ├── linker.js           ← bidirectional wikilink writer
│   ├── frontmatter.js      ← YAML generator (preserves existing user fields)
│   └── detectors/
│       ├── role.js           ← 3-tier role detection
│       ├── capabilities.js   ← 14 capability patterns
│       ├── techniques.js     ← 12 prompt-engineering technique patterns
│       ├── tools.js          ← 10 tool patterns
│       ├── domain.js         ← content-based domain detection
│       └── sophistication.js ← weighted scorer
├── cli/index.js            ← mindgrove analyze <path>
├── sample-vault/           ← 6 ready-to-run sample markdown files
├── examples/
│   └── ai-prompt-vault.js  ← real-world 1300-file use case
└── obsidian/               ← original Obsidian Templater script (inspiration)

License

MIT © Kevork @ Nexacrawl