feat: add AI docs assistant knowledge base (E0, E1)

- Add 6 FAQ markdown files for clientbeheer, intake, screening,
  behandelplan, spraak, and inloggen
- Add 2 guidelines files for interface and technical documentation
- Implement CategoryDetector for keyword-based category detection
- Implement KnowledgeLoader for loading relevant knowledge files
- Implement PromptBuilder for combining base prompt with knowledge

Part of AI Documentatie Assistent feature (Epic 0 & Epic 1)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

lib/docs/knowledge-loader.ts

@@ -0,0 +1,97 @@
+import fs from 'fs/promises'
+import path from 'path'
+
+import { KnowledgeCategory } from '@/lib/docs/category-detector'
+
+const KNOWLEDGE_DIR = path.join(process.cwd(), 'lib/docs/knowledge')
+
+const CATEGORY_FILE_MAP: Record<KnowledgeCategory, string[]> = {
+  clientbeheer: ['faq_clientbeheer.md'],
+  intake: ['faq_intake.md'],
+  screening: ['faq_screening.md'],
+  behandelplan: ['faq_behandelplan.md'],
+  spraak: ['faq_spraak.md'],
+  inloggen: ['faq_inloggen.md'],
+  interface: ['guidelines_interface.md'],
+  technisch: ['guidelines_technisch.md'],
+}
+
+const FILE_CATEGORY_MAP = Object.entries(CATEGORY_FILE_MAP).reduce(
+  (acc, [category, files]) => {
+    for (const file of files) {
+      acc[file] = category as KnowledgeCategory
+    }
+    return acc
+  },
+  {} as Record<string, KnowledgeCategory>
+)
+
+const DEFAULT_FILES = ['guidelines_interface.md']
+const MAX_SECTIONS = 4
+
+export interface KnowledgeSection {
+  id: string
+  title: string
+  content: string
+  category?: KnowledgeCategory
+  file: string
+}
+
+const normalizeTitle = (fileName: string) =>
+  fileName
+    .replace(/_/g, ' ')
+    .replace(/\.mdx?$/, '')
+    .trim()
+
+function extractTitle(content: string, fallback: string) {
+  const match = content.match(/^#\s+(.+)$/m)
+  return (match ? match[1] : fallback).trim()
+}
+
+async function readKnowledgeFile(fileName: string) {
+  const filePath = path.join(KNOWLEDGE_DIR, fileName)
+  const fileContent = await fs.readFile(filePath, 'utf-8')
+  return fileContent.trim()
+}
+
+export async function loadKnowledgeSections(
+  categories: KnowledgeCategory[],
+  options?: { limit?: number; fallbackFiles?: string[] }
+): Promise<KnowledgeSection[]> {
+  const limit = options?.limit ?? MAX_SECTIONS
+  const fallbackFiles = options?.fallbackFiles ?? DEFAULT_FILES
+
+  const selectedFiles = new Set<string>()
+
+  if (categories.length === 0) {
+    fallbackFiles.forEach(file => selectedFiles.add(file))
+  } else {
+    for (const category of categories) {
+      const files = CATEGORY_FILE_MAP[category] ?? []
+      files.forEach(file => selectedFiles.add(file))
+    }
+  }
+
+  const filesToLoad = Array.from(selectedFiles).slice(0, limit)
+  const sections: KnowledgeSection[] = []
+
+  for (const fileName of filesToLoad) {
+    try {
+      const raw = await readKnowledgeFile(fileName)
+      if (!raw) continue
+
+      const title = extractTitle(raw, normalizeTitle(fileName))
+      sections.push({
+        id: fileName.replace(/\.mdx?$/, ''),
+        title,
+        content: raw,
+        category: FILE_CATEGORY_MAP[fileName],
+        file: fileName,
+      })
+    } catch (error) {
+      console.error(`Failed to load knowledge file ${fileName}:`, error)
+    }
+  }
+
+  return sections
+}