bacchus-labs/codebase-index

Codebase Index

Generate comprehensive, framework-agnostic relationship and dependency maps for component-based codebases with optional TOON format for token-efficient AI consumption.

What This Skill Does

Scans any component-based codebase to automatically generate relationship maps:

• Component relationships: Import graph (usedBy/uses)
• NPM dependencies: Package usage tracking
• Utility functions: Shared code in lib/utils
• Data queries: CMS and API patterns
• CSS structure: Classes and design tokens

Framework Support: React, Vue, Svelte, Astro, Next.js, Angular, Solid (auto-detected)

Output Formats:

• TOON (default): Token-efficient format, 30-60% smaller than JSON
• JSON: Standard format for tooling compatibility

Quick Start

# Auto-detect framework, output TOON
python scripts/index_codebase.py /path/to/project

# Specify framework and format
python scripts/index_codebase.py /path/to/project --framework react --format json

# Use tab delimiters (even more token-efficient)
python scripts/index_codebase.py /path/to/project --delimiter "\t"

Framework Auto-Detection

The skill automatically detects your framework from package.json:

| Framework | Detection | Extensions | |-----------|-----------|------------| | Astro | astro dependency | .astro | | React | react dependency | .jsx, .tsx | | Next.js | next dependency | .jsx, .tsx | | Vue | vue dependency | .vue | | Svelte | svelte dependency | .svelte | | Angular | @angular/core | .ts | | Solid | solid-js | .jsx, .tsx |

Override auto-detection:

python scripts/index_codebase.py /path/to/project --framework vue

Output Formats

TOON Format (Recommended)

Why TOON?

• 30-60% fewer tokens than JSON
• Better LLM comprehension (70% vs 65% accuracy)
• Self-documenting structure with explicit lengths
• Perfect for uniform data (our use case!)

Example TOON output:

components[3]{name,path,type,framework,uses,usedBy}:
Button,src/components/atoms/Button.tsx,atom,react,[0]:,[2]: Card,Page
Card,src/components/molecules/Card.tsx,molecule,react,[1]: Button,[1]: Page
Page,src/pages/index.tsx,page,react,[2]: Button,Card,[0]:

statistics:
  totalComponents: 3
  atoms: 1
  molecules: 1
  pages: 1

vs JSON (same data):

{
  "components": {
    "Button": {
      "path": "src/components/atoms/Button.tsx",
      "type": "atom",
      "framework": "react",
      "uses": [],
      "usedBy": ["Card", "Page"]
    },
    ...
  }
}

Format Comparison

| Aspect | TOON | JSON | |--------|------|------| | Token count | 40-70% smaller | Baseline | | LLM accuracy | 70.1% | 65.4% | | Human readable | ✅ | ✅ | | Tool compatible | ⚠️ (new format) | ✅ | | Best for | AI consumption | Tooling |

Recommendation: Use TOON for AI context, JSON for tooling integration.

Output Structure

All files generated in src/.ai/ (or first source directory found):

src/.ai/
├── index.toon                      # Entry point
└── relationships/
    ├── component-usage.toon        # Component graph
    ├── dependencies.toon           # NPM + utilities + CSS
    └── data-flow.toon              # Data queries

Component Type Detection

Auto-detects component types from directory structure:

| Pattern | Type | |---------|------| | /atoms/ | atom | | /molecules/ | molecule | | /organisms/ | organism | | /templates/ | template | | /ui/ | ui | | /layouts/ | layout | | /pages/, /views/, /routes/ | page | | /hooks/ | hook | | /contexts/ | context | | Other | component |

Advanced Options

Alternative Delimiters

For tabular data, choose delimiter for optimal tokenization:

# Comma (default) - most readable
python scripts/index_codebase.py /path/to/project --delimiter ","

# Tab - often fewer tokens
python scripts/index_codebase.py /path/to/project --delimiter "\t"

# Pipe - middle ground
python scripts/index_codebase.py /path/to/project --delimiter "|"

Tab output example:

components[2	]{name	path	type}:
Button	src/Button.tsx	atom
Card	src/Card.tsx	molecule

Dual Output

Generate both formats for different use cases:

# Generate TOON for AI
python scripts/index_codebase.py /path/to/project --format toon

# Generate JSON for tools
python scripts/index_codebase.py /path/to/project --format json

Reading Index Files

For AI Assistants

After indexing, read files to understand codebase:

1. Start with index.toon/json - Overview and metadata
2. Component relationships - relationships/component-usage.toon
3. Dependencies - relationships/dependencies.toon
4. Data flow - relationships/data-flow.toon

TOON Syntax Quick Reference

# Primitive field
name: Button

# Nested object
user:
  id: 123
  name: Ada

# Primitive array
tags[3]: react,typescript,ui

# Tabular array (uniform objects)
items[2]{id,name,price}:
1,Widget,9.99
2,Gadget,14.50

# List array (mixed)
items[3]:
- 1
- name: Item
- [2]: sub,array

Framework-Specific Features

Astro

• Detects .astro components
• Scans for Sanity CMS client.fetch() queries
• Maps layouts and pages

React/Next.js

• Supports .jsx, .tsx
• Detects hooks, contexts
• Maps fetch() and axios patterns

Vue

• Scans .vue components
• Maps views and composables

Svelte

• Detects .svelte components
• Scans routes and stores

Angular

• Scans .component.ts files
• Detects http.get() patterns

Solid

• Supports .jsx, .tsx
• Maps Solid-specific patterns

Project Structure Requirements

Flexible directory discovery:

• Checks framework-specific directories (e.g., app/ for Next.js)
• Falls back to src/ if present
• Works with various project structures

Typical structures supported:

# React/Next.js
app/
components/
pages/

# Vue
src/
  components/
  views/

# Astro
src/
  components/
  layouts/
  pages/

Integration with Metadata

Complementary Design

Component Metadata (manual .metadata.json):

• Purpose: "How to USE this component"
• Contains: Props, variants, examples, guidelines
• Created: Manually

Codebase Index (this skill):

• Purpose: "WHERE is this used"
• Contains: Relationships, dependencies, usage
• Created: Auto-generated

No duplication - These systems complement each other.

Performance

• Speed: 1-5 seconds for typical projects (50-200 components)
• Output size:
  • TOON: 30-150KB (typical)
  • JSON: 50-250KB (typical)
• Memory: Minimal (streaming processing)

Large projects (500+ components):

• Consider directory-specific scans
• Use tab delimiter for better tokenization
• May take 10-30 seconds

Error Handling

The indexer is resilient:

• Skips unparseable files (logs warning)
• Handles missing directories gracefully
• Continues on errors to maximize coverage
• Reports errors in summary

Example Usage

First-time Index

# Let it auto-detect everything
python scripts/index_codebase.py ~/projects/my-app

# Output:
# 🔍 Scanning react codebase...
# ✅ Generated component-usage.toon
# ✅ Generated dependencies.toon
# ✅ Generated data-flow.toon
# ✅ Generated index.toon
#
# Summary:
# 🎯 Framework: react
# 📝 Format: TOON
# ✅ 42 components indexed
# ✅ 156 relationships mapped
# ✅ 8 utilities tracked
# ✅ 3 queries documented
# 💡 TOON format: 30-60% more token-efficient than JSON

Re-index After Changes

# Same command - overwrites previous index
python scripts/index_codebase.py ~/projects/my-app

Framework Override

# For edge cases where auto-detect fails
python scripts/index_codebase.py ~/projects/my-app --framework vue

Troubleshooting

No components found?

• Verify source directories exist
• Check file extensions match framework
• Look for node_modules in exclusions

Wrong framework detected?

• Use --framework flag to override
• Check package.json dependencies

Import relationships missing?

• Verify import statements use standard ES6 format
• Dynamic imports may not be detected
• Check for path aliases (may need resolution)

TOON parse errors?

• File is valid - TOON is for AI consumption
• Use JSON format for tooling integration
• Check TOON spec if implementing parser

Utilities not found?

• Verify lib/, utils/, or helpers/ directories exist
• Check for .ts files (not .d.ts)

Comparison with JSON

Example: Component Data

JSON (123 tokens):

{
  "Button": {
    "path": "src/components/Button.tsx",
    "type": "atom",
    "framework": "react",
    "uses": [],
    "usedBy": ["Card", "Header", "Footer"]
  }
}

TOON (52 tokens, 58% reduction):

Button:
  path: src/components/Button.tsx
  type: atom
  framework: react
  uses[0]:
  usedBy[3]: Card,Header,Footer

Tabular TOON (even better for multiple components):

components[3]{name,path,type,uses,usedBy}:
Button,src/components/Button.tsx,atom,[0]:,[3]: Card,Header,Footer
Card,src/components/Card.tsx,molecule,[1]: Button,[1]: Page
Footer,src/components/Footer.tsx,organism,[2]: Button,Icon,[0]:

CI/CD Integration

Git Hook Example

# .husky/post-merge
#!/bin/sh
python scripts/index_codebase.py . --format toon
git add src/.ai/

GitHub Action

name: Update Codebase Index
on:
  push:
    branches: [main]
jobs:
  index:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      - name: Index codebase
        run: python scripts/index_codebase.py . --format toon
      - name: Commit changes
        run: |
          git add src/.ai/
          git commit -m "Update codebase index" || true
          git push

Future Enhancements

Potential additions:

• Bundle size analysis
• Circular dependency detection
• Visual graph generation (mermaid)
• More CMS platforms (Contentful, Strapi)
• API route mapping
• Environment variable tracking
• Incremental updates

Credits

• TOON Format: toon-format/toon
• Token Efficiency: 30-60% reduction vs JSON
• LLM Accuracy: 70.1% vs 65.4% (JSON baseline)