troglodyte/
├── src/
│   ├── index.ts              # Entry point
│   ├── promptPreprocessor.ts # Pipeline orchestrator + system metadata extraction
│   ├── troglodyte.ts         # Compression engine (main logic)
│   ├── config.ts             # UI configuration schematics
│   └── dictionaries/
│       ├── en-filler.ts      # English blacklists (gentle/balanced/aggressive)
│       ├── de-filler.ts      # German blacklists
│       ├── phrases.ts        # Multi-word phrase replacements
│       └── synonyms.ts       # Single-word abbreviations
├── dist/                     # Compiled output
├── package.json
└── tsconfig.json

// Before: verbose placeholders like "\uE001P1\uE001" = 7+ chars
const PU = '\uE001';
generatePlaceholder() => `${PU}P${++counter}${PU}`;

// After: compact single-char placeholders = 2 chars
String.fromCodePoint(0xE000 + (counter++ % 0xFFF)); // "\uE000"

C:\Source Code\ServiceMonitor\...
↓ synonym replacement
C:\src Code\ServiceMonitor\...  ❌ CORRUPTED!

text = text.replace(/([A-Za-z]:[\/\\][^<>"|?*\r\n]{10,})/g, (match) => {
  return protectIfWorthwhile(match, 15);
});

const sortedPhrases = Object.entries(this.phrasesAndLogic)
  .sort((a, b) => b[0].length - a[0].length);

text = result
  .replace(/\s+/g, ' ')                           // Collapse spaces
  .replace(/\s+([.,?!;:])/g, '$1')                // Remove space BEFORE punct
  .replace(/([.?!;:])(?=[A-ZßÄÖÜ])/g, '$1 ')      // Add space AFTER (before CAPITAL)
  .trim();

// BEFORE (BROKEN)
text = text.replace(/[-￿]/g, ...);

// AFTER (FIXED)
text = text.replace(/[\uE000-\uF8FF]/g, ...);

const MAX_PLACEHOLDERS = 0xFFFFF; // ~1 million
if (placeholderCounter >= MAX_PLACEHOLDERS) {
  console.warn('[Troglodyte] ⚠️ Placeholder limit reached!');
  return match; // Skip protection, return original
}

const errorMessage = error instanceof Error ? error.message : String(error);
console.error("[Troglodyte] Stack trace:", error.stack);
status.setState({ text: `Compression failed (${errorMessage.substring(0, 40)}...)` });

// Relative paths FIRST (before absolute to prevent /lib from being consumed)
text = text.replace(/(\.\.?\/[^\s<>"|?*]+)(?=[$\s.,;:!?)\]]|$)/g, protectIfWorthwhile);

// Then absolute paths with `/` included in character class
text = text.replace(/(\/[^\s<>"|?*]+)(?=[$\s.,;:!?)\]]|$)/g, protectIfWorthwhile);

const enIndicators = ['the', 'a', ...]; // Array
if (enIndicators.includes(word)) ...     // O(n) per word

const enIndicators = new Set(['the', 'a', ...]); // Set
if (enIndicators.has(word)) ...           // O(1) per word

interface CompiledPhrase {
  phrase: string;
  replacement: string | undefined;
  regex: RegExp; // Compiled ONCE!
}

let result = '';
for (...) { result += token; }  // O(n²)

const parts: string[] = [];
for (...) { parts.push(token); }
let result = parts.join('');     // O(n)

for (let i = 0; i < protectedItems.length; i++) {
  text = text.split(placeholder).join(item); // O(n) per iteration
}

const replacements = new Map();
for (let i = 0; i < protectedItems.length; i++) {
  replacements.set(String.fromCodePoint(0xE000 + i), protectedItems[i]);
}
text = text.replace(/[-￿]/g, (match) => 
  replacements.get(match) || match
); // O(n) single pass!

const wordPattern = /[a-zA-Z0-9_.\-'ßäöüÄÖÜ]+/g;  // Now includes .

Input:  "Hallo! Ich würde mich sehr freuen,"
After filtering "Hallo", "Ich", "mich", "sehr": "!  würde   freuen,"
                                                ↑        ↑
                                         orphaned !  orphaned ,

.replace(/\s+([.,?!;:])\s+/g, ' ')  // Remove standalone punctuation → single space

if (!prompt || typeof prompt !== 'string') {
  console.warn('[Troglodyte] Invalid input: prompt must be a non-empty string');
  return prompt || '';
}

const MAX_INPUT_LENGTH = 1_000_000; // 1MB limit to prevent DoS
if (prompt.length > MAX_INPUT_LENGTH) {
  console.warn(`[Troglodyte] Input exceeds ${MAX_INPUT_LENGTH} char limit...`);
}

return str.replace(/[.*+?^${}()|[\]\\]/g, '\\---


## 📊 Performance Metrics');

Input: "Hello there! I was wondering if you could possibly help me out? 
        I would really appreciate it if you could explain how to install 
        Node.js on Windows step by step. Thank you so much for your time 
        and assistance!"

Output: "possibly help me out? explain how install Node.js Windows steps. 
         and assistance."

Ratio: ~65-70% compression ✅

Input: "check C:\Source Code\ServiceMonitor:\ServiceMonitor \ for issues."

Output: "check C:\Source Code\ServiceMonitor:\ServiceMonitor \ for issues."

Path preserved intact ✅ (not corrupted to "C:\src Code\...")

Input: "Bitte analysiere den Code in /home/user/project/src/main.ts und ./lib/utils.py."

Output: "analysiere Code /home/user/project/src/main.ts ./lib/utils.py."

Both paths preserved intact ✅ (extensions included, no fragmentation)

cd "C:\Source Code\LM Studio Plugins\troglodyte"
lms dev --install

npm run dev

return totalTokens > 0 && (codeScore / totalTokens) > 0.15; // was 0.25

const protectedPronouns = new Set([
  // English
  'he', 'him', 'his', 'she', 'her', 'it', 'they', 'them', 'their',
  // German
  'er', 'ihn', 'ihm', 'sein', 'sie', 'ihr', 'es', 'wir', 'uns', 'euch', 'mein', 'dein',
]);

'step by step': 'sequential',  // instead of 'steps'
'Schritt für Schritt': 'sequenziell',  // German equivalent

return totalTokens > 0 && (codeScore / totalTokens) > 0.25; // was 0.1

.replace(/[^a-zA-Z0-9äöüßÄÖÜ]+$/, ''); // Strip mixed punctuation
if (['?', '!'].includes(lastChar)) { text = text + lastChar; } // Re-add if original ended in ? or !

const tokenPattern = /([^\s\w]+)|(\w+)/gu; // Captures delimiters (group 1) & words (group 2)
for (const m of text.matchAll(tokenPattern)) { /* process */ }

const pattern = /\b(?:const|let|var|function|class|import|export)\b|[{}]/g;
while ((match = pattern.exec(text)) !== null) codeScore++;

const WORD_TOKEN_REGEX = /\b[a-zäöüß]{3,}\b/g; // Hoisted

this.synonymMap = new Map(Object.entries(dictionaries.synonyms || {}));
const replacement = this.synonymMap.get(lower); // O(1) vs Object property access

this.emptyReplacements = new Set();
for (const val of this.replacementMap.values()) {
  if (!val || !val.trim()) this.emptyReplacements.add(val!);
}
// In hot loop:
if (this.emptyReplacements.has(repl)) return ' '; // No .trim() call!

troglodyte/
├── src/
│   ├── index.ts              # Entry point
│   ├── promptPreprocessor.ts # Pipeline orchestrator + system metadata extraction
│   ├── troglodyte.ts         # Compression engine (main logic)
│   ├── config.ts             # UI configuration schematics
│   └── dictionaries/
│       ├── en-filler.ts      # English blacklists (gentle/balanced/aggressive)
│       ├── de-filler.ts      # German blacklists
│       ├── phrases.ts        # Multi-word phrase replacements
│       └── synonyms.ts       # Single-word abbreviations
├── dist/                     # Compiled output
├── package.json
└── tsconfig.json

// Before: verbose placeholders like "\uE001P1\uE001" = 7+ chars
const PU = '\uE001';
generatePlaceholder() => `${PU}P${++counter}${PU}`;

// After: compact single-char placeholders = 2 chars
String.fromCodePoint(0xE000 + (counter++ % 0xFFF)); // "\uE000"

C:\Source Code\ServiceMonitor\...
↓ synonym replacement
C:\src Code\ServiceMonitor\...  ❌ CORRUPTED!

text = text.replace(/([A-Za-z]:[\/\\][^<>"|?*\r\n]{10,})/g, (match) => {
  return protectIfWorthwhile(match, 15);
});

const sortedPhrases = Object.entries(this.phrasesAndLogic)
  .sort((a, b) => b[0].length - a[0].length);

text = result
  .replace(/\s+/g, ' ')                           // Collapse spaces
  .replace(/\s+([.,?!;:])/g, '$1')                // Remove space BEFORE punct
  .replace(/([.?!;:])(?=[A-ZßÄÖÜ])/g, '$1 ')      // Add space AFTER (before CAPITAL)
  .trim();

// BEFORE (BROKEN)
text = text.replace(/[-￿]/g, ...);

// AFTER (FIXED)
text = text.replace(/[\uE000-\uF8FF]/g, ...);

const MAX_PLACEHOLDERS = 0xFFFFF; // ~1 million
if (placeholderCounter >= MAX_PLACEHOLDERS) {
  console.warn('[Troglodyte] ⚠️ Placeholder limit reached!');
  return match; // Skip protection, return original
}

const errorMessage = error instanceof Error ? error.message : String(error);
console.error("[Troglodyte] Stack trace:", error.stack);
status.setState({ text: `Compression failed (${errorMessage.substring(0, 40)}...)` });

// Relative paths FIRST (before absolute to prevent /lib from being consumed)
text = text.replace(/(\.\.?\/[^\s<>"|?*]+)(?=[$\s.,;:!?)\]]|$)/g, protectIfWorthwhile);

// Then absolute paths with `/` included in character class
text = text.replace(/(\/[^\s<>"|?*]+)(?=[$\s.,;:!?)\]]|$)/g, protectIfWorthwhile);

const enIndicators = ['the', 'a', ...]; // Array
if (enIndicators.includes(word)) ...     // O(n) per word

const enIndicators = new Set(['the', 'a', ...]); // Set
if (enIndicators.has(word)) ...           // O(1) per word

interface CompiledPhrase {
  phrase: string;
  replacement: string | undefined;
  regex: RegExp; // Compiled ONCE!
}

let result = '';
for (...) { result += token; }  // O(n²)

const parts: string[] = [];
for (...) { parts.push(token); }
let result = parts.join('');     // O(n)

for (let i = 0; i < protectedItems.length; i++) {
  text = text.split(placeholder).join(item); // O(n) per iteration
}

const replacements = new Map();
for (let i = 0; i < protectedItems.length; i++) {
  replacements.set(String.fromCodePoint(0xE000 + i), protectedItems[i]);
}
text = text.replace(/[-￿]/g, (match) => 
  replacements.get(match) || match
); // O(n) single pass!

const wordPattern = /[a-zA-Z0-9_.\-'ßäöüÄÖÜ]+/g;  // Now includes .

Input:  "Hallo! Ich würde mich sehr freuen,"
After filtering "Hallo", "Ich", "mich", "sehr": "!  würde   freuen,"
                                                ↑        ↑
                                         orphaned !  orphaned ,

.replace(/\s+([.,?!;:])\s+/g, ' ')  // Remove standalone punctuation → single space

if (!prompt || typeof prompt !== 'string') {
  console.warn('[Troglodyte] Invalid input: prompt must be a non-empty string');
  return prompt || '';
}

const MAX_INPUT_LENGTH = 1_000_000; // 1MB limit to prevent DoS
if (prompt.length > MAX_INPUT_LENGTH) {
  console.warn(`[Troglodyte] Input exceeds ${MAX_INPUT_LENGTH} char limit...`);
}

return str.replace(/[.*+?^${}()|[\]\\]/g, '\\---


## 📊 Performance Metrics');

Input: "Hello there! I was wondering if you could possibly help me out? 
        I would really appreciate it if you could explain how to install 
        Node.js on Windows step by step. Thank you so much for your time 
        and assistance!"

Output: "possibly help me out? explain how install Node.js Windows steps. 
         and assistance."

Ratio: ~65-70% compression ✅

Input: "check C:\Source Code\ServiceMonitor:\ServiceMonitor \ for issues."

Output: "check C:\Source Code\ServiceMonitor:\ServiceMonitor \ for issues."

Path preserved intact ✅ (not corrupted to "C:\src Code\...")

Input: "Bitte analysiere den Code in /home/user/project/src/main.ts und ./lib/utils.py."

Output: "analysiere Code /home/user/project/src/main.ts ./lib/utils.py."

Both paths preserved intact ✅ (extensions included, no fragmentation)

cd "C:\Source Code\LM Studio Plugins\troglodyte"
lms dev --install

npm run dev

return totalTokens > 0 && (codeScore / totalTokens) > 0.15; // was 0.25

const protectedPronouns = new Set([
  // English
  'he', 'him', 'his', 'she', 'her', 'it', 'they', 'them', 'their',
  // German
  'er', 'ihn', 'ihm', 'sein', 'sie', 'ihr', 'es', 'wir', 'uns', 'euch', 'mein', 'dein',
]);

'step by step': 'sequential',  // instead of 'steps'
'Schritt für Schritt': 'sequenziell',  // German equivalent

return totalTokens > 0 && (codeScore / totalTokens) > 0.25; // was 0.1

.replace(/[^a-zA-Z0-9äöüßÄÖÜ]+$/, ''); // Strip mixed punctuation
if (['?', '!'].includes(lastChar)) { text = text + lastChar; } // Re-add if original ended in ? or !

const tokenPattern = /([^\s\w]+)|(\w+)/gu; // Captures delimiters (group 1) & words (group 2)
for (const m of text.matchAll(tokenPattern)) { /* process */ }

const pattern = /\b(?:const|let|var|function|class|import|export)\b|[{}]/g;
while ((match = pattern.exec(text)) !== null) codeScore++;

const WORD_TOKEN_REGEX = /\b[a-zäöüß]{3,}\b/g; // Hoisted

this.synonymMap = new Map(Object.entries(dictionaries.synonyms || {}));
const replacement = this.synonymMap.get(lower); // O(1) vs Object property access

this.emptyReplacements = new Set();
for (const val of this.replacementMap.values()) {
  if (!val || !val.trim()) this.emptyReplacements.add(val!);
}
// In hot loop:
if (this.emptyReplacements.has(repl)) return ' '; // No .trim() call!