feat: Implement AI-powered estimation system, add comprehensive usage guide, and redesign PDF sitemap module.

scripts/ai-estimate.ts

@@ -23,6 +23,8 @@ async function main() {
 
     let jsonStatePath: string | null = null;
 
+    const isEstimation = process.argv.includes('--estimation') || process.argv.includes('-E');
+
     const args = process.argv.slice(2);
     for (let i = 0; i < args.length; i++) {
         const arg = args[i];
@@ -34,6 +36,8 @@ async function main() {
             cacheKey = args[++i];
         } else if (arg === '--json') {
             jsonStatePath = args[++i];
+        } else if (arg === '--estimation' || arg === '-E') {
+            // Handled above
         } else if (!arg.startsWith('--')) {
             briefing = arg;
         }
@@ -140,7 +144,8 @@ async function main() {
     console.log(`📦 Saved detailed state to: ${finalJsonPath}`);
     console.log('📄 Generating PDF estimation...');
     try {
-        execSync(`npx tsx ./scripts/generate-quote.ts --input ${tempJsonPath}`, { stdio: 'inherit' });
+        const genArgs = isEstimation ? '--estimation' : '';
+        execSync(`npx tsx ./scripts/generate-quote.ts --input ${tempJsonPath} ${genArgs}`, { stdio: 'inherit' });
     } finally {
         // await fs.unlink(tempJsonPath);
     }
@@ -243,7 +248,17 @@ async function performCrawl(url: string): Promise<string> {
     return summary + pages.map(p => `--- PAGE: ${p.url} ---\n${p.content}`).join('\n\n');
 }
 
-async function getAiEstimation(briefing: string, distilledCrawl: string, comments: string | null, apiKey: string, principles: string, techStandards: string, tone: string) {
+const cleanJson = (str: string) => {
+    // Remove markdown code blocks if present
+    let cleaned = str.replace(/```json\n?|```/g, '').trim();
+    // Remove potential control characters that break JSON.parse
+    cleaned = cleaned.replace(/[\u0000-\u001F\u007F-\u009F]/g, " ");
+    // Remove trailing commas before closing braces/brackets
+    cleaned = cleaned.replace(/,\s*([\]}])/g, '$1');
+    return cleaned;
+};
+
+const getAiEstimation = async (briefing: string, distilledCrawl: string, comments: string | null, apiKey: string, principles: string, techStandards: string, tone: string) => {
     let usage = { prompt: 0, completion: 0, cost: 0 };
     const addUsage = (data: any) => {
         if (data?.usage) {
@@ -316,7 +331,7 @@ Focus 100% on the BRIEFING text provided by the user. Use the DISTILLED_CRAWL on
         response_format: { type: 'json_object' }
     }, { headers: { 'Authorization': `Bearer ${apiKey}`, 'Content-Type': 'application/json' } });
     addUsage(p1Resp.data);
-    const facts = JSON.parse(p1Resp.data.choices[0].message.content);
+    const facts = JSON.parse(cleanJson(p1Resp.data.choices[0].message.content));
 
     // 2. PASS 2: Feature Deep-Dive
     console.log('  ↳ Pass 2: Feature Deep-Dive...');
@@ -348,7 +363,7 @@ ${JSON.stringify(facts, null, 2)}
         response_format: { type: 'json_object' }
     }, { headers: { 'Authorization': `Bearer ${apiKey}`, 'Content-Type': 'application/json' } });
     addUsage(p2Resp.data);
-    const details = JSON.parse(p2Resp.data.choices[0].message.content);
+    const details = JSON.parse(cleanJson(p2Resp.data.choices[0].message.content));
 
     // 3. PASS 3: Strategic Content (Bespoke Strategy)
     console.log('  ↳ Pass 3: Strategic Content (Bespoke Strategy)...');
@@ -360,19 +375,17 @@ Analyze the BRIEFING and the EXISTING WEBSITE context.
 ${tone}
 
 ### OBJECTIVE:
-1. **briefingSummary**: A deep, respectful summary of the status quo and the target state.
-   - **LENGTH**: EXACTLY TWO PARAGRAPHS. Minimum 8 sentences total.
-   - **MIRROR TEST**: Acknowledge the EXISTING website specifically. Why does the new project make sense NOW?
-   - **ABSOLUTE RULE**: DO NOT claim "keine digitale Repräsentation", "erstmals abgebildet", "Erstplatzierung" or "kommunikative Lücke" regarding existence if Pass 1 identified this as a RELAUNCH (isRelaunch=true). EXPLICITLY acknowledge the existing context and the NEED FOR EVOLUTION/MODERNIZATION.
-   - **RESPECT**: Explicitly incorporate the customer's expressed wishes.
-   - **ABSOLUTE RULE**: DO NOT INVENT DETAILS. Do not mention specific people (e.g., "Frieder Helmich"), software versions, or internal details NOT present in the briefing. 
-   - **TONE**: Natural Ich-Form. Clear, direct, zero marketing fluff.
-2. **designVision**: A high-density, professional vision of the future execution.
-   - **LENGTH**: EXACTLY TWO PARAGRAPHS. Minimum 6 sentences total.
-   - **INVESTMENT VALUE**: Plant a clear picture of a stable, high-quality system.
-   - **TECHNICAL PRECISION**: Focus on execution (Typografie, Visual Logic, Performance).
-   - **NO FLUFF**: Do NOT focus on "Full-Screen Hero Video" as the main thing. Focus on the FIRM's essence and how we translate it into a professional tool.
-   - **ABSOLUTE RULE**: NO HALLUCINATIONS. Stay general yet precise. No "Verzicht auf Stockmaterial" unless explicitly stated.
+3. **briefingSummary**: Ein sachlicher, tiefgehender Überblick der Unternehmenslage.
+   - **STIL**: Keine Ich-Form. Keine Marketing-Floskeln. Nutze präzise Fachbegriffe. Sei prägnant und effizient (ca. 70% der vorherigen Länge).
+   - **FORM**: EXAKT ZWEI ABSÄTZE. Insgesamt ca. 6 Sätze.
+   - **INHALT**: Welcher technologische Sprung ist notwendig? Was ist der Status Quo? (Bezug zur URL/Briefing).
+   - **ABSOLUTE REGEL**: Keine Halluzinationen über fehlende Präsenzen bei Relaunches.
+   - **DATENSCHUTZ**: KEINERLEI namentliche Nennungen von Personen (z. B. "Danny Joseph") in diesen Texten.
+4. **designVision**: Ein abstraktes, strategisches Konzept.
+   - **STIL**: Rein konzeptionell. Keine Umsetzungsschritte. Keinerlei "To-dos". Sei prägnant.
+   - **FORM**: EXAKT ZWEI ABSÄTZE. Insgesamt ca. 4 Sätze.
+   - **DATENSCHUTZ**: KEINERLEI namentliche Nennungen von Personen in diesen Texten.
+   - **FOKUS**: Welche strategische Wirkung soll erzielt werden? (Z. B. "Industrielle Souveränität").
 
 ### OUTPUT FORMAT (Strict JSON):
 {
@@ -389,7 +402,7 @@ ${tone}
         response_format: { type: 'json_object' }
     }, { headers: { 'Authorization': `Bearer ${apiKey}`, 'Content-Type': 'application/json' } });
     addUsage(p3Resp.data);
-    const strategy = JSON.parse(p3Resp.data.choices[0].message.content);
+    const strategy = JSON.parse(cleanJson(p3Resp.data.choices[0].message.content));
 
     // 4. PASS 4: Information Architecture (Sitemap)
     console.log('  ↳ Pass 4: Information Architecture...');
@@ -418,7 +431,7 @@ ${JSON.stringify({ facts, strategy }, null, 2)}
         response_format: { type: 'json_object' }
     }, { headers: { 'Authorization': `Bearer ${apiKey}`, 'Content-Type': 'application/json' } });
     addUsage(p4Resp.data);
-    const ia = JSON.parse(p4Resp.data.choices[0].message.content);
+    const ia = JSON.parse(cleanJson(p4Resp.data.choices[0].message.content));
 
     // 5. PASS 5: Position Synthesis & Pricing Transparency
     console.log('  ↳ Pass 5: Position Synthesis...');
@@ -443,14 +456,15 @@ Each position in the quote must be perfectly justified and detailed.
 ### RULES FOR positionDescriptions (STRICT):
 1. **NO "ICH-FORM"**: Do NOT use "Ich" or "Mein". Lead with the action or component.
 2. **CONCISE & ITEM-BASED**: Use short, technical sentences. Focus on WHAT is delivered.
-3. **ZERO GENERALIZATION**: Do NOT say "Verschiedene Funktionen".
+3. **ZERO GENERALIZATION**: Do NOT say "Verschiedene Funktionen" or "Optimierte Darstellung". Name the things.
 4. **ITEMIZED SYNTHESIS**: Mention EVERY component selected in Pass 1. 
-5. **HARD SPECIFICS**: Preserve technical details from the briefing (e.g., "110 kV", "HDD-Bohrtechnik").
-6. **STYLE**: Direct, engineering-grade, 100% GERMAN.
+5. **HARD SPECIFICS**: Preserve technical details from the briefing and CURRENT WEBSITE (distilledCrawl). If they mention "HDD-Bohrtechnik" or "110kV Kabelanlagen", IT MUST BE IN THE DESCRIPTION.
+6. **INDUSTRIAL AMBITION**: Describe it as a high-end technical solution, not a cheap website.
 7. **SPECIFIC - PAGES**: For "Individuelle Seiten", list the pages as a comma-separated list.
 8. **SPECIFIC - HOSTING**: Always append: "Inkl. 20GB Speicher."
 9. **SPECIFIC - LOGIC**: Describe the ACTUAL logic. NEVER use generic terms.
 10. **STRICT KEYS**: Keys MUST be EXACTLY the ones defined in POSITION TITLES.
+11. **AGB BAN (CRITICAL)**: NEVER mention "Allgemeine Geschäftsbedingungen" or "AGGs" in any text. They are NOT part of this offer.
 
 ### EXAMPLES (FEW-SHOT):
 - **BAD**: "Ich entwickle die Seiten: Startseite, Leistungen, Kontakt."
@@ -481,7 +495,7 @@ ${JSON.stringify({ facts, details, strategy, ia }, null, 2)}
         response_format: { type: 'json_object' }
     }, { headers: { 'Authorization': `Bearer ${apiKey}`, 'Content-Type': 'application/json' } });
     addUsage(p5Resp.data);
-    const positionsData = JSON.parse(p5Resp.data.choices[0].message.content);
+    const positionsData = JSON.parse(cleanJson(p5Resp.data.choices[0].message.content));
 
     // 6. PASS 6: The Industrial Critic
     console.log('  ↳ Pass 6: The Industrial Critic (Quality Gate)...');
@@ -492,10 +506,13 @@ Analyze the CURRENT_STATE against the BRIEFING_TRUTH.
 ### CRITICAL ERROR CHECKLIST (FAIL IF FOUND):
 1. **Hallucination Leakage**: FAIL if names of people (e.g., "Frieder Helmich"), specific software versions, or invented details are used unless they appear EXACTLY in the BRIEFING.
    - **CRITICAL**: Forbid "Sie", "Ansprechpartner" or "Unternehmen" for personName if a name IS in the briefing. If none is in briefing, use empty string.
-2. **Logic Conflict**: FAIL if isRelaunch is true but briefingSummary claims no website exists or uses phrases like "Da aktuell keine digitale Repräsentation vorliegt", "erstmals abgebildet", "Erstplatzierung" or "Lücke schließen" (regarding existence).
+2. **Logic Conflict**: FAIL if isRelaunch is true but briefingSummary claims no website exists.
    - FAIL if the description in positionDescriptions mentions more items than extracted in facts.
-3. **Implementation Fluff**: FAIL if "React", "Next.js", "TypeScript", "Tailwind" or other tech-stack details are mentioned. Focus on Concept & Result.
-4. **Length Check**: Briefing and Vision MUST be significantly long (EXACTLY 2 paragraphs each, minimum 8 sentences for briefing, 6 for vision).
+3. **Implementation Fluff**: FAIL if tech-stack details are mentioned (React, etc.). Focus on Concept & Result.
+4. **Genericism Check (CRITICAL)**: FAIL if any text sounds like it could apply to ANY company. It MUST mention specific industry details (e.g., "Kabeltiefbau", "Infrastruktur-Zentrum") from the Briefing or Crawl.
+6. **Namen-Verbot (STRICT)**: FAIL if any personal names (e.g. "Danny Joseph", "Joseph", etc.) appear in 'briefingSummary' or 'designVision'. Use abstract terms like "Unternehmensführung" or "Management" if necessary.
+7. **AGB BAN**: FAIL if "Allgemeine Geschäftsbedingungen" or "AGB" appear anywhere.
+8. **Length Check**: Briefing (ca. 6 Sätze) und Vision (ca. 4 Sätze). Kürze Texte, die zu ausschweifend sind, auf das Wesentliche.
 
 ### MISSION:
 Return updated fields ONLY. Specifically focus on hardening 'positionDescriptions', 'sitemap', 'briefingSummary', and 'designVision'.
@@ -509,7 +526,7 @@ ${JSON.stringify({ facts, strategy, ia, positionsData }, null, 2)}
         response_format: { type: 'json_object' }
     }, { headers: { 'Authorization': `Bearer ${apiKey}`, 'Content-Type': 'application/json' } });
     addUsage(p6Resp.data);
-    const reflection = JSON.parse(p6Resp.data.choices[0].message.content);
+    const reflection = JSON.parse(cleanJson(p6Resp.data.choices[0].message.content));
 
     // 6. Reflection Merge Utility
     const mergeReflection = (state: any, reflection: any) => {