Phase 1 — Python (klausur-service): 5 monoliths → 36 files - dsfa_corpus_ingestion.py (1,828 LOC → 5 files) - cv_ocr_engines.py (2,102 LOC → 7 files) - cv_layout.py (3,653 LOC → 10 files) - vocab_worksheet_api.py (2,783 LOC → 8 files) - grid_build_core.py (1,958 LOC → 6 files) Phase 2 — Go (edu-search-service, school-service): 8 monoliths → 19 files - staff_crawler.go (1,402 → 4), policy/store.go (1,168 → 3) - policy_handlers.go (700 → 2), repository.go (684 → 2) - search.go (592 → 2), ai_extraction_handlers.go (554 → 2) - seed_data.go (591 → 2), grade_service.go (646 → 2) Phase 3 — TypeScript (admin-lehrer): 45 monoliths → 220+ files - sdk/types.ts (2,108 → 16 domain files) - ai/rag/page.tsx (2,686 → 14 files) - 22 page.tsx files split into _components/ + _hooks/ - 11 component files split into sub-components - 10 SDK data catalogs added to loc-exceptions - Deleted dead backup index_original.ts (4,899 LOC) All original public APIs preserved via re-export facades. Zero new errors: Python imports verified, Go builds clean, TypeScript tsc --noEmit shows only pre-existing errors. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
144 lines
5.1 KiB
TypeScript
144 lines
5.1 KiB
TypeScript
'use client'
|
|
|
|
import { useEffect, useRef, useState, useMemo } from 'react'
|
|
import type { GridCell } from '@/app/(admin)/ai/ocr-kombi/types'
|
|
import type { WordPosition } from './usePixelWordPositions'
|
|
|
|
interface LlmReviewOverlayProps {
|
|
cells: GridCell[]
|
|
imageNaturalSize: { w: number; h: number } | null
|
|
fontScale: number
|
|
leftPaddingPct: number
|
|
globalBold: boolean
|
|
cellWordPositions: Map<string, WordPosition[]>
|
|
}
|
|
|
|
export function LlmReviewOverlay({
|
|
cells,
|
|
imageNaturalSize,
|
|
fontScale,
|
|
leftPaddingPct,
|
|
globalBold,
|
|
cellWordPositions,
|
|
}: LlmReviewOverlayProps) {
|
|
const reconRef = useRef<HTMLDivElement>(null)
|
|
const [reconWidth, setReconWidth] = useState(0)
|
|
|
|
// Track reconstruction container width for font size calculation
|
|
useEffect(() => {
|
|
const el = reconRef.current
|
|
if (!el) return
|
|
const obs = new ResizeObserver(entries => {
|
|
for (const entry of entries) setReconWidth(entry.contentRect.width)
|
|
})
|
|
obs.observe(el)
|
|
return () => obs.disconnect()
|
|
}, [])
|
|
|
|
// Snap all cells in the same column to consistent x/w positions
|
|
const colPositions = useMemo(() => {
|
|
const byCol = new Map<number, { xs: number[]; ws: number[] }>()
|
|
for (const cell of cells) {
|
|
if (!cell.bbox_pct) continue
|
|
const entry = byCol.get(cell.col_index) || { xs: [], ws: [] }
|
|
entry.xs.push(cell.bbox_pct.x)
|
|
entry.ws.push(cell.bbox_pct.w)
|
|
byCol.set(cell.col_index, entry)
|
|
}
|
|
const result = new Map<number, { x: number; w: number }>()
|
|
for (const [colIdx, { xs, ws }] of byCol) {
|
|
xs.sort((a, b) => a - b)
|
|
ws.sort((a, b) => a - b)
|
|
const medianX = xs[Math.floor(xs.length / 2)]
|
|
const medianW = ws[Math.floor(ws.length / 2)]
|
|
result.set(colIdx, { x: medianX, w: medianW })
|
|
}
|
|
return result
|
|
}, [cells])
|
|
|
|
return (
|
|
<div>
|
|
<div className="text-xs font-medium text-gray-500 dark:text-gray-400 mb-1">
|
|
Text-Rekonstruktion ({cells.filter(c => c.text).length} Zellen)
|
|
</div>
|
|
<div className="border border-gray-200 dark:border-gray-700 rounded-lg overflow-hidden bg-white dark:bg-white">
|
|
<div
|
|
ref={reconRef}
|
|
className="relative"
|
|
style={{
|
|
aspectRatio: imageNaturalSize ? `${imageNaturalSize.w} / ${imageNaturalSize.h}` : '3 / 4',
|
|
}}
|
|
>
|
|
{cells.map(cell => {
|
|
if (!cell.bbox_pct || !cell.text) return null
|
|
const col = colPositions.get(cell.col_index)
|
|
const cellX = col?.x ?? cell.bbox_pct.x
|
|
const cellW = col?.w ?? cell.bbox_pct.w
|
|
const aspect = imageNaturalSize ? imageNaturalSize.h / imageNaturalSize.w : 4 / 3
|
|
const containerH = reconWidth * aspect
|
|
const cellHeightPx = containerH * (cell.bbox_pct.h / 100)
|
|
|
|
const wordPos = cellWordPositions.get(cell.cell_id)
|
|
|
|
// Pixel-analysed: render word-groups at detected positions
|
|
if (wordPos) {
|
|
return wordPos.map((wp, i) => {
|
|
const autoFontPx = cellHeightPx * wp.fontRatio * fontScale
|
|
const fs = Math.max(6, autoFontPx)
|
|
return (
|
|
<span
|
|
key={`${cell.cell_id}_${i}`}
|
|
className="absolute leading-none pointer-events-none select-none"
|
|
style={{
|
|
left: `${wp.xPct}%`,
|
|
top: `${cell.bbox_pct.y}%`,
|
|
width: `${wp.wPct}%`,
|
|
height: `${cell.bbox_pct.h}%`,
|
|
fontSize: `${fs}px`,
|
|
fontWeight: globalBold ? 'bold' : (cell.is_bold ? 'bold' : 'normal'),
|
|
fontFamily: "'Liberation Sans', Arial, sans-serif",
|
|
display: 'flex',
|
|
alignItems: 'center',
|
|
whiteSpace: 'nowrap',
|
|
overflow: 'visible',
|
|
color: '#1a1a1a',
|
|
}}
|
|
>
|
|
{wp.text}
|
|
</span>
|
|
)
|
|
})
|
|
}
|
|
|
|
// Fallback: no pixel data - single span for entire cell
|
|
const fontSize = Math.max(6, cellHeightPx * fontScale)
|
|
return (
|
|
<span
|
|
key={cell.cell_id}
|
|
className="absolute leading-none pointer-events-none select-none"
|
|
style={{
|
|
left: `${cellX}%`,
|
|
top: `${cell.bbox_pct.y}%`,
|
|
width: `${cellW}%`,
|
|
height: `${cell.bbox_pct.h}%`,
|
|
fontSize: `${fontSize}px`,
|
|
fontWeight: globalBold ? 'bold' : (cell.is_bold ? 'bold' : 'normal'),
|
|
paddingLeft: `${leftPaddingPct}%`,
|
|
fontFamily: "'Liberation Sans', Arial, sans-serif",
|
|
display: 'flex',
|
|
alignItems: 'center',
|
|
whiteSpace: 'pre',
|
|
overflow: 'visible',
|
|
color: '#1a1a1a',
|
|
}}
|
|
>
|
|
{cell.text}
|
|
</span>
|
|
)
|
|
})}
|
|
</div>
|
|
</div>
|
|
</div>
|
|
)
|
|
}
|