Files
breakpilot-lehrer/admin-lehrer/components/ocr-pipeline/StepColumnDetection.tsx
Benjamin Admin 23b7840ea7
Some checks failed
CI / go-lint (push) Has been skipped
CI / python-lint (push) Has been skipped
CI / nodejs-lint (push) Has been skipped
CI / test-go-school (push) Successful in 40s
CI / test-go-edu-search (push) Successful in 28s
CI / test-python-klausur (push) Failing after 2m16s
CI / test-python-agent-core (push) Successful in 17s
CI / test-nodejs-website (push) Successful in 22s
feat: Full-Row OCR mit Spacing fuer Box-Sub-Sessions
Sub-Sessions ueberspringen Spaltenerkennung und nutzen stattdessen eine
Pseudo-Spalte ueber die volle Breite. Text wird mit proportionalem
Spacing aus Wort-Positionen rekonstruiert, um raeumliches Layout zu erhalten.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-10 08:28:29 +01:00

433 lines
16 KiB
TypeScript

'use client'
import { useCallback, useEffect, useState } from 'react'
import type { ColumnResult, ColumnGroundTruth, PageRegion, SubSession } from '@/app/(admin)/ai/ocr-pipeline/types'
import { ColumnControls } from './ColumnControls'
import { ManualColumnEditor } from './ManualColumnEditor'
import type { ColumnTypeKey } from '@/app/(admin)/ai/ocr-pipeline/types'
const KLAUSUR_API = '/klausur-api'
type ViewMode = 'normal' | 'ground-truth' | 'manual'
interface StepColumnDetectionProps {
sessionId: string | null
onNext: () => void
onBoxSessionsCreated?: (subSessions: SubSession[]) => void
}
/** Convert PageRegion[] to divider percentages + column types for ManualColumnEditor */
function columnsToEditorState(
columns: PageRegion[],
imageWidth: number
): { dividers: number[]; columnTypes: ColumnTypeKey[] } {
if (!columns.length || !imageWidth) return { dividers: [], columnTypes: [] }
const sorted = [...columns].sort((a, b) => a.x - b.x)
const dividers: number[] = []
const columnTypes: ColumnTypeKey[] = sorted.map(c => c.type)
for (let i = 1; i < sorted.length; i++) {
const xPct = (sorted[i].x / imageWidth) * 100
dividers.push(xPct)
}
return { dividers, columnTypes }
}
export function StepColumnDetection({ sessionId, onNext, onBoxSessionsCreated }: StepColumnDetectionProps) {
const [columnResult, setColumnResult] = useState<ColumnResult | null>(null)
const [detecting, setDetecting] = useState(false)
const [error, setError] = useState<string | null>(null)
const [viewMode, setViewMode] = useState<ViewMode>('normal')
const [applying, setApplying] = useState(false)
const [imageDimensions, setImageDimensions] = useState<{ width: number; height: number } | null>(null)
const [savedGtColumns, setSavedGtColumns] = useState<PageRegion[] | null>(null)
const [creatingBoxSessions, setCreatingBoxSessions] = useState(false)
const [existingSubSessions, setExistingSubSessions] = useState<SubSession[] | null>(null)
const [isSubSession, setIsSubSession] = useState(false)
// Fetch session info (image dimensions) + check for cached column result
useEffect(() => {
if (!sessionId || imageDimensions) return
const fetchSessionInfo = async () => {
try {
const infoRes = await fetch(`${KLAUSUR_API}/api/v1/ocr-pipeline/sessions/${sessionId}`)
if (infoRes.ok) {
const info = await infoRes.json()
if (info.image_width && info.image_height) {
setImageDimensions({ width: info.image_width, height: info.image_height })
}
const isSub = !!info.parent_session_id
setIsSubSession(isSub)
if (info.sub_sessions && info.sub_sessions.length > 0) {
setExistingSubSessions(info.sub_sessions)
onBoxSessionsCreated?.(info.sub_sessions)
}
if (info.column_result) {
setColumnResult(info.column_result)
// Sub-session with pseudo-column already set → auto-advance
if (isSub) {
onNext()
return
}
return
}
// Sub-session without columns → auto-detect (creates pseudo-column)
if (isSub) {
const res = await fetch(`${KLAUSUR_API}/api/v1/ocr-pipeline/sessions/${sessionId}/columns`, { method: 'POST' })
if (res.ok) {
const data: ColumnResult = await res.json()
setColumnResult(data)
onNext()
return
}
}
}
} catch (e) {
console.error('Failed to fetch session info:', e)
}
// No cached result - run auto-detection
runAutoDetection()
}
fetchSessionInfo()
// eslint-disable-next-line react-hooks/exhaustive-deps
}, [sessionId])
// Load saved GT if exists
useEffect(() => {
if (!sessionId) return
const fetchGt = async () => {
try {
const res = await fetch(`${KLAUSUR_API}/api/v1/ocr-pipeline/sessions/${sessionId}/ground-truth/columns`)
if (res.ok) {
const data = await res.json()
const corrected = data.columns_gt?.corrected_columns
if (corrected) setSavedGtColumns(corrected)
}
} catch {
// No saved GT - that's fine
}
}
fetchGt()
}, [sessionId])
const runAutoDetection = useCallback(async () => {
if (!sessionId) return
setDetecting(true)
setError(null)
try {
const res = await fetch(`${KLAUSUR_API}/api/v1/ocr-pipeline/sessions/${sessionId}/columns`, {
method: 'POST',
})
if (!res.ok) {
const err = await res.json().catch(() => ({ detail: res.statusText }))
throw new Error(err.detail || 'Spaltenerkennung fehlgeschlagen')
}
const data: ColumnResult = await res.json()
setColumnResult(data)
} catch (e) {
setError(e instanceof Error ? e.message : 'Unbekannter Fehler')
} finally {
setDetecting(false)
}
}, [sessionId])
const handleRerun = useCallback(() => {
runAutoDetection()
}, [runAutoDetection])
const handleGroundTruth = useCallback(async (gt: ColumnGroundTruth) => {
if (!sessionId) return
try {
await fetch(`${KLAUSUR_API}/api/v1/ocr-pipeline/sessions/${sessionId}/ground-truth/columns`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify(gt),
})
} catch (e) {
console.error('Ground truth save failed:', e)
}
}, [sessionId])
const handleManualApply = useCallback(async (columns: PageRegion[]) => {
if (!sessionId) return
setApplying(true)
setError(null)
try {
const res = await fetch(`${KLAUSUR_API}/api/v1/ocr-pipeline/sessions/${sessionId}/columns/manual`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ columns }),
})
if (!res.ok) {
const err = await res.json().catch(() => ({ detail: res.statusText }))
throw new Error(err.detail || 'Manuelle Spalten konnten nicht gespeichert werden')
}
const data = await res.json()
setColumnResult({
columns: data.columns,
duration_seconds: data.duration_seconds ?? 0,
})
setViewMode('normal')
} catch (e) {
setError(e instanceof Error ? e.message : 'Fehler beim Speichern')
} finally {
setApplying(false)
}
}, [sessionId])
const handleGtApply = useCallback(async (columns: PageRegion[]) => {
if (!sessionId) return
setApplying(true)
setError(null)
try {
const gt: ColumnGroundTruth = {
is_correct: false,
corrected_columns: columns,
}
await fetch(`${KLAUSUR_API}/api/v1/ocr-pipeline/sessions/${sessionId}/ground-truth/columns`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify(gt),
})
setSavedGtColumns(columns)
setViewMode('normal')
} catch (e) {
setError(e instanceof Error ? e.message : 'Fehler beim Speichern')
} finally {
setApplying(false)
}
}, [sessionId])
// Count box zones from column result
const boxZones = columnResult?.zones?.filter(z => z.zone_type === 'box') || []
const boxCount = boxZones.length
const createBoxSessions = useCallback(async () => {
if (!sessionId) return
setCreatingBoxSessions(true)
setError(null)
try {
const res = await fetch(`${KLAUSUR_API}/api/v1/ocr-pipeline/sessions/${sessionId}/create-box-sessions`, {
method: 'POST',
})
if (!res.ok) {
const err = await res.json().catch(() => ({ detail: res.statusText }))
throw new Error(err.detail || 'Box-Sessions konnten nicht erstellt werden')
}
const data = await res.json()
const subs: SubSession[] = data.sub_sessions.map((s: { id: string; name?: string; box_index: number }) => ({
id: s.id,
name: s.name || `Box ${s.box_index + 1}`,
box_index: s.box_index,
current_step: 1,
status: 'pending',
}))
setExistingSubSessions(subs)
onBoxSessionsCreated?.(subs)
} catch (e) {
setError(e instanceof Error ? e.message : 'Fehler beim Erstellen der Box-Sessions')
} finally {
setCreatingBoxSessions(false)
}
}, [sessionId, onBoxSessionsCreated])
if (!sessionId) {
return (
<div className="flex flex-col items-center justify-center py-16 text-center">
<div className="text-5xl mb-4">📊</div>
<h3 className="text-lg font-medium text-gray-700 dark:text-gray-300 mb-2">
Schritt 3: Spaltenerkennung
</h3>
<p className="text-gray-500 dark:text-gray-400 max-w-md">
Bitte zuerst Schritt 1 und 2 abschliessen.
</p>
</div>
)
}
const dewarpedUrl = `${KLAUSUR_API}/api/v1/ocr-pipeline/sessions/${sessionId}/image/cropped`
const overlayUrl = `${KLAUSUR_API}/api/v1/ocr-pipeline/sessions/${sessionId}/image/columns-overlay`
// Pre-compute editor state from saved GT or auto columns for GT mode
const gtInitial = savedGtColumns
? columnsToEditorState(savedGtColumns, imageDimensions?.width ?? 1000)
: undefined
return (
<div className="space-y-4">
{/* Loading indicator */}
{detecting && (
<div className="flex items-center gap-2 text-teal-600 dark:text-teal-400 text-sm">
<div className="animate-spin w-4 h-4 border-2 border-teal-500 border-t-transparent rounded-full" />
Spaltenerkennung laeuft...
</div>
)}
{viewMode === 'manual' ? (
/* Manual column editor - overwrites column_result */
<ManualColumnEditor
imageUrl={dewarpedUrl}
imageWidth={imageDimensions?.width ?? 1000}
imageHeight={imageDimensions?.height ?? 1400}
onApply={handleManualApply}
onCancel={() => setViewMode('normal')}
applying={applying}
mode="manual"
/>
) : viewMode === 'ground-truth' ? (
/* GT mode: auto result (left, readonly) + GT editor (right) */
<div className="grid grid-cols-2 gap-4">
{/* Left: Auto result (readonly overlay) */}
<div>
<div className="text-xs font-medium text-gray-500 dark:text-gray-400 mb-1">
Auto-Ergebnis (readonly)
</div>
<div className="border rounded-lg overflow-hidden dark:border-gray-700 bg-gray-50 dark:bg-gray-900">
{columnResult ? (
// eslint-disable-next-line @next/next/no-img-element
<img
src={`${overlayUrl}?t=${Date.now()}`}
alt="Auto Spalten-Overlay"
className="w-full h-auto"
/>
) : (
<div className="aspect-[3/4] flex items-center justify-center text-gray-400 text-sm">
Keine Auto-Daten
</div>
)}
</div>
{/* Auto column list */}
{columnResult && (
<div className="mt-2 space-y-1">
<div className="text-xs font-medium text-gray-500 dark:text-gray-400">
Auto: {columnResult.columns.length} Spalten
</div>
{columnResult.columns
.filter(c => c.type.startsWith('column') || c.type === 'page_ref')
.map((col, i) => (
<div key={i} className="text-xs text-gray-500 dark:text-gray-400 font-mono">
{i + 1}. {col.type} x={col.x} w={col.width}
</div>
))}
</div>
)}
</div>
{/* Right: GT editor */}
<div>
<div className="text-xs font-medium text-gray-500 dark:text-gray-400 mb-1">
Ground Truth Editor
</div>
<ManualColumnEditor
imageUrl={dewarpedUrl}
imageWidth={imageDimensions?.width ?? 1000}
imageHeight={imageDimensions?.height ?? 1400}
onApply={handleGtApply}
onCancel={() => setViewMode('normal')}
applying={applying}
mode="ground-truth"
layout="stacked"
initialDividers={gtInitial?.dividers}
initialColumnTypes={gtInitial?.columnTypes}
/>
</div>
</div>
) : (
/* Normal mode: overlay (left) vs clean (right) */
<div className="grid grid-cols-2 gap-4">
<div>
<div className="text-xs font-medium text-gray-500 dark:text-gray-400 mb-1">
Mit Spalten-Overlay
</div>
<div className="border rounded-lg overflow-hidden dark:border-gray-700 bg-gray-50 dark:bg-gray-900">
{columnResult ? (
// eslint-disable-next-line @next/next/no-img-element
<img
src={`${overlayUrl}?t=${Date.now()}`}
alt="Spalten-Overlay"
className="w-full h-auto"
/>
) : (
<div className="aspect-[3/4] flex items-center justify-center text-gray-400 text-sm">
{detecting ? 'Erkenne Spalten...' : 'Keine Daten'}
</div>
)}
</div>
</div>
<div>
<div className="text-xs font-medium text-gray-500 dark:text-gray-400 mb-1">
Entzerrtes Bild
</div>
<div className="border rounded-lg overflow-hidden dark:border-gray-700 bg-gray-50 dark:bg-gray-900">
{/* eslint-disable-next-line @next/next/no-img-element */}
<img
src={dewarpedUrl}
alt="Entzerrt"
className="w-full h-auto"
/>
</div>
</div>
</div>
)}
{/* Box zone info */}
{viewMode === 'normal' && boxCount > 0 && (
<div className="bg-amber-50 dark:bg-amber-900/20 border border-amber-200 dark:border-amber-700 rounded-xl p-4 flex items-center justify-between">
<div className="flex items-center gap-3">
<span className="text-2xl">📦</span>
<div>
<div className="text-sm font-medium text-amber-800 dark:text-amber-300">
{boxCount} Box{boxCount > 1 ? 'en' : ''} erkannt
</div>
<div className="text-xs text-amber-600 dark:text-amber-400">
Box-Bereiche werden separat verarbeitet
</div>
</div>
</div>
{existingSubSessions && existingSubSessions.length > 0 ? (
<div className="text-xs text-amber-700 dark:text-amber-300 font-medium">
{existingSubSessions.length} Box-Session{existingSubSessions.length > 1 ? 's' : ''} vorhanden
</div>
) : (
<button
onClick={createBoxSessions}
disabled={creatingBoxSessions}
className="px-4 py-2 bg-amber-600 text-white rounded-lg hover:bg-amber-700 transition-colors text-sm font-medium disabled:opacity-50 flex items-center gap-2"
>
{creatingBoxSessions && (
<div className="animate-spin w-3.5 h-3.5 border-2 border-white border-t-transparent rounded-full" />
)}
Box-Sessions erstellen
</button>
)}
</div>
)}
{/* Controls */}
{viewMode === 'normal' && (
<ColumnControls
columnResult={columnResult}
onRerun={handleRerun}
onManualMode={() => setViewMode('manual')}
onGtMode={() => setViewMode('ground-truth')}
onGroundTruth={handleGroundTruth}
onNext={onNext}
isDetecting={detecting}
savedGtColumns={savedGtColumns}
/>
)}
{error && (
<div className="p-3 bg-red-50 dark:bg-red-900/20 text-red-600 dark:text-red-400 rounded-lg text-sm">
{error}
</div>
)}
</div>
)
}