fix(personal): complete scanned PDF OCR gate

This commit is contained in:
2026-07-12 22:08:02 +08:00
parent faf070e171
commit fa73d413c5
5 changed files with 120 additions and 5 deletions
@@ -185,7 +185,18 @@ public class PersonalIngestionWorker {
try {
StoredObject stored = tenantRunner.execute(item.tenantId(), () -> objectReader.read(
item.ossId(), ownerObjectPrefix(item), item.ownerUserId(), maxInputBytes));
ParsedDocument document = parser.parse(stored.fileName(), item.mimeType(), stored.bytes());
ParsedDocument document;
try {
document = parser.parse(stored.fileName(), item.mimeType(), stored.bytes());
} catch (KnowledgeDocumentParser.ParseException exception) {
if (exception.failure() == KnowledgeDocumentParser.Failure.EMPTY && isPdf(item, stored)) {
pdfOcrHandler.enqueue(new PersonalPdfOcrService.OcrItem(item.id(), item.tenantId(),
item.spaceId(), item.ownerUserId(), item.ossId(), item.tagsJson(), attemptVersion),
stored.bytes());
return true;
}
throw exception;
}
List<String> chunks = document.chunks(chunkSize, chunkOverlap);
if (chunks.isEmpty()) {
if (isPdf(item, stored)) {
@@ -352,14 +352,21 @@ public class PersonalPdfOcrService {
where tenant_id = ? and owner_user_id = ? and item_id = ? and status = 'FAILED'
order by page_number
""", Integer.class, job.tenantId(), job.ownerUserId(), job.itemId());
List<String> failureCodes = pageTexts.isEmpty() ? jdbcTemplate.queryForList("""
select distinct last_error_code from aihr_personal_ocr_page
where tenant_id = ? and owner_user_id = ? and item_id = ?
and status = 'FAILED' and last_error_code is not null
order by last_error_code
""", String.class, job.tenantId(), job.ownerUserId(), job.itemId()) : List.of();
transactions.execute(status -> {
if (pageTexts.isEmpty()) {
String failureCode = terminalFailureCode(failureCodes);
jdbcTemplate.update("""
update aihr_personal_item
set status = 'FAILED', error_code = 'PERSONAL_OCR_EMPTY',
error_message = '扫描PDF未识别到可用文字', update_time = now()
set status = 'FAILED', error_code = ?, error_message = ?, update_time = now()
where tenant_id = ? and owner_user_id = ? and id = ? and status = 'PARSING'
""", job.tenantId(), job.ownerUserId(), job.itemId());
""", failureCode, terminalFailureMessage(failureCode),
job.tenantId(), job.ownerUserId(), job.itemId());
} else {
publishFragments(job, String.join("\n\n", pageTexts), failedPages);
}
@@ -374,6 +381,26 @@ public class PersonalPdfOcrService {
});
}
private static String terminalFailureCode(List<String> failureCodes) {
if (failureCodes.size() == 1 && !"PERSONAL_OCR_PAGE_EMPTY".equals(failureCodes.get(0))) {
return failureCodes.get(0);
}
if (failureCodes.stream().anyMatch(code -> code != null && !"PERSONAL_OCR_PAGE_EMPTY".equals(code))) {
return "PERSONAL_OCR_UNAVAILABLE";
}
return "PERSONAL_OCR_EMPTY";
}
private static String terminalFailureMessage(String failureCode) {
return switch (failureCode) {
case "PERSONAL_OCR_MODEL_UNAVAILABLE" -> "未配置支持图片的视觉模型,请配置后重试失败页";
case "PERSONAL_OCR_RUNTIME_DISABLED" -> "扫描PDF识别服务已关闭,请启用后重试失败页";
case "PERSONAL_OCR_PROVIDER_FAILED", "PERSONAL_OCR_RESPONSE_INVALID", "PERSONAL_OCR_UNAVAILABLE" ->
"扫描PDF识别服务暂不可用,请稍后重试失败页";
default -> "扫描PDF未识别到可用文字";
};
}
private void publishFragments(OcrJob job, String text, List<Integer> failedPages) {
List<String> chunks = new ParsedDocument(text, "application/pdf", Map.of()).chunks(800, 120);
jdbcTemplate.update("""
@@ -44,7 +44,8 @@ class PersonalIngestionWorkerTest {
when(jdbc.queryForList(contains("status = 'QUEUED'"))).thenReturn(List.of(item()));
when(jdbc.update(contains("status = 'PARSING'"), eq("000000"), eq(101L), eq(9L), eq(0))).thenReturn(1);
when(parser.parse(eq("sample.pdf"), eq("text/plain"), any(byte[].class)))
.thenReturn(new ParsedDocument("", "application/pdf", Map.of()));
.thenThrow(new KnowledgeDocumentParser.ParseException(
KnowledgeDocumentParser.Failure.EMPTY, "document contains no text"));
AtomicReference<org.dromara.aihr.personal.service.PersonalPdfOcrService.OcrItem> handedOff =
new AtomicReference<>();
PersonalIngestionWorker worker = PersonalIngestionWorker.forTestWithPdfOcr(
@@ -165,6 +165,34 @@ class PersonalPdfOcrServiceTest {
eq("000000"), eq(101L), eq(9L));
}
@Test
void allPagesRejectedByProviderExposeModelFailureInsteadOfEmptyDocument() {
JdbcTemplate jdbc = mock(JdbcTemplate.class);
Map<String, Object> job = Map.ofEntries(
Map.entry("id", 5L), Map.entry("tenant_id", "000000"), Map.entry("owner_user_id", 101L),
Map.entry("item_id", 9L), Map.entry("space_id", 7L), Map.entry("oss_id", 88L),
Map.entry("tags_json", "[]"), Map.entry("total_pages", 8), Map.entry("next_page", 9),
Map.entry("status", "RETRY"));
when(jdbc.queryForList(contains("from aihr_personal_ocr_job"))).thenReturn(List.of(job));
when(jdbc.update(contains("set status = 'RUNNING'"), eq(5L))).thenReturn(1);
when(jdbc.queryForList(contains("attempt_count < 3"), eq(Integer.class),
eq("000000"), eq(101L), eq(9L), eq(20))).thenReturn(List.of());
when(jdbc.queryForList(contains("select ocr_text"), eq(String.class),
eq("000000"), eq(101L), eq(9L))).thenReturn(List.of());
when(jdbc.queryForList(contains("select page_number"), eq(Integer.class),
eq("000000"), eq(101L), eq(9L))).thenReturn(List.of(1, 2, 3, 4, 5, 6, 7, 8));
when(jdbc.queryForList(contains("select distinct last_error_code"), eq(String.class),
eq("000000"), eq(101L), eq(9L))).thenReturn(List.of("PERSONAL_OCR_PROVIDER_FAILED"));
PersonalPdfOcrService service = PersonalPdfOcrService.forTest(
jdbc, mock(PersonalPdfPageRenderer.class), mock(PersonalVisionOcrService.class),
immediateTransactions(), ignored -> new byte[] {1});
assertTrue(service.processNextBatch());
verify(jdbc).update(contains("set status = 'FAILED', error_code = ?"),
eq("PERSONAL_OCR_PROVIDER_FAILED"), any(String.class), eq("000000"), eq(101L), eq(9L));
}
@Test
void recoversExpiredRunningLease() {
JdbcTemplate jdbc = mock(JdbcTemplate.class);