fix(ocr): unblock event loop during OCR and show errors in UI

OCR engines are CPU-bound and were blocking Uvicorn's single async
event loop, making /health unresponsive during processing. This caused
new OCR requests to fail silently (health check failure → no DB record
→ UI shows NONE). Wrap engine calls in asyncio.to_thread() to keep the
event loop free. Also surface OCR trigger errors in the frontend
instead of silently resetting the spinner.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Marcel
2026-04-12 23:50:39 +02:00
parent ef11e4af09
commit d8dcba1a71
2 changed files with 19 additions and 2 deletions

View File

@@ -7,6 +7,7 @@ import TranscriptionEditView from '$lib/components/TranscriptionEditView.svelte'
import TranscriptionReadView from '$lib/components/TranscriptionReadView.svelte';
import TranscriptionPanelHeader from '$lib/components/TranscriptionPanelHeader.svelte';
import type { TranscriptionBlockData } from '$lib/types';
import { getErrorMessage } from '$lib/errors';
let { data } = $props();
@@ -129,6 +130,7 @@ async function reviewToggle(blockId: string) {
let ocrRunning = $state(false);
let ocrProgressMessage = $state('');
let ocrErrorMessage = $state('');
let ocrPollTimer = $state<ReturnType<typeof setInterval> | null>(null);
function translateOcrProgress(code: string): string {
@@ -154,6 +156,7 @@ function translateOcrProgress(code: string): string {
async function triggerOcr(scriptType: string) {
ocrRunning = true;
ocrErrorMessage = '';
try {
const res = await fetch(`/api/documents/${doc.id}/ocr`, {
method: 'POST',
@@ -165,10 +168,14 @@ async function triggerOcr(scriptType: string) {
pollOcrJob(data.jobId);
} else {
ocrRunning = false;
const body = await res.json().catch(() => null);
const code = (body as { code?: string } | null)?.code;
ocrErrorMessage = code ? getErrorMessage(code) : m.ocr_status_error();
}
} catch (e) {
console.error('Failed to trigger OCR:', e);
ocrRunning = false;
ocrErrorMessage = m.ocr_status_error();
}
}
@@ -185,6 +192,9 @@ function pollOcrJob(jobId: string) {
ocrPollTimer = null;
ocrRunning = false;
ocrProgressMessage = '';
if (job.status === 'FAILED') {
ocrErrorMessage = m.ocr_status_error();
}
await loadTranscriptionBlocks();
annotationReloadKey++;
panelMode = transcriptionBlocks.length > 0 ? 'read' : 'edit';
@@ -399,6 +409,11 @@ onMount(() => {
onClose={() => (transcribeMode = false)}
/>
<div class="flex-1 overflow-y-auto">
{#if ocrErrorMessage}
<div class="mx-4 mt-4 rounded-sm border border-red-200 bg-red-50 px-4 py-3">
<p class="text-sm text-red-700">{ocrErrorMessage}</p>
</div>
{/if}
{#if ocrRunning}
<div class="flex flex-1 flex-col items-center justify-center px-6 py-12 text-center">
<svg

View File

@@ -1,5 +1,6 @@
"""OCR microservice — FastAPI app with Surya and Kraken engine support."""
import asyncio
import io
import logging
from contextlib import asynccontextmanager
@@ -52,6 +53,7 @@ async def run_ocr(request: OcrRequest):
Downloads the PDF from the provided URL, converts pages to images,
and runs the appropriate OCR engine based on scriptType.
OCR engines run in a thread pool so the event loop stays free for /health.
"""
if not _models_ready:
raise HTTPException(status_code=503, detail="Models not loaded yet")
@@ -66,10 +68,10 @@ async def run_ocr(request: OcrRequest):
status_code=400,
detail="Kraken model not available — cannot process Kurrent script",
)
blocks = kraken_engine.extract_blocks(images, request.language)
blocks = await asyncio.to_thread(kraken_engine.extract_blocks, images, request.language)
else:
# TYPEWRITER, HANDWRITING_LATIN, UNKNOWN — all use Surya
blocks = surya_engine.extract_blocks(images, request.language)
blocks = await asyncio.to_thread(surya_engine.extract_blocks, images, request.language)
threshold = get_threshold(script_type)
for block in blocks: