Skip to content

RAGFlow is an agent-native database: a Retrieval-Augmented Generation (RAG) engine that grew into context infrastructure.

fix: surface MinerU API error bodies and re-raise parser failures (#17308)

天海蒼灆a1208cd2733d
diff --git a/deepdoc/parser/mineru_parser.py b/deepdoc/parser/mineru_parser.py
index 17c54447ba..b77212eef3 100644
--- a/deepdoc/parser/mineru_parser.py
+++ b/deepdoc/parser/mineru_parser.py
@@ -297,12 +297,11 @@ class MinerUParser(RAGFlowPdfParser):
data = {
"output_dir": "./output",
- "lang_list": options.lang,
- "backend": options.backend,
- "parse_method": options.method,
+ "lang_list": options.lang.value if isinstance(options.lang, MinerULanguage) else options.lang,
+ "backend": options.backend.value if isinstance(options.backend, MinerUBackend) else options.backend,
+ "parse_method": options.method.value if isinstance(options.method, MinerUParseMethod) else options.method,
"formula_enable": options.formula_enable,
"table_enable": options.table_enable,
- "server_url": None,
"return_md": True,
"return_middle_json": True,
"return_model_output": True,
@@ -324,7 +323,7 @@ class MinerUParser(RAGFlowPdfParser):
headers = {"Accept": "application/json"}
try:
- self.logger.info(f"[MinerU] invoke api: {self.mineru_api}/file_parse backend={options.backend} server_url={data.get('server_url')}")
+ self.logger.info(f"[MinerU] invoke api: {self.mineru_api}/file_parse backend={data['backend']} server_url={data.get('server_url')}")
if callback:
callback(0.20, f"[MinerU] invoke api: {self.mineru_api}/file_parse")
with open(pdf_file_path, "rb") as pdf_file:
@@ -337,7 +336,9 @@ class MinerUParser(RAGFlowPdfParser):
timeout=1800,
stream=True,
) as response:
- response.raise_for_status()
+ if not response.ok:
+ body = (response.text or "")[:2000]
+ raise RuntimeError(f"[MinerU] api failed status={response.status_code} body={body}")
content_type = response.headers.get("Content-Type", "")
if not content_type.startswith("application/zip"):
raise RuntimeError(f"[MinerU] not zip returned from api: {content_type}")
@@ -354,7 +355,14 @@ class MinerUParser(RAGFlowPdfParser):
self.logger.info("[MinerU] Api completed successfully.")
return Path(output_path)
except requests.RequestException as e:
- raise RuntimeError(f"[MinerU] api failed with exception {e}")
+ detail = ""
+ resp = getattr(e, "response", None)
+ if resp is not None:
+ try:
+ detail = f" body={(resp.text or '')[:2000]}"
+ except Exception:
+ pass
+ raise RuntimeError(f"[MinerU] api failed with exception {e}{detail}") from e
def __images__(self, fnm, zoomin: int = 1, page_from=0, page_to=MAXIMUM_PAGE_NUMBER, callback=None):
self.page_from = page_from
diff --git a/rag/app/naive.py b/rag/app/naive.py
index 3f4c8498ab..612279d7a2 100644
--- a/rag/app/naive.py
+++ b/rag/app/naive.py
@@ -413,6 +413,9 @@ def by_paddleocr(
return sections, tables, pdf_parser
except Exception as e:
logging.error(f"Failed to parse pdf via LLMBundle PaddleOCR ({paddleocr_llm_name}): {e}")
+ if callback:
+ callback(-1, f"Failed to parse pdf via PaddleOCR ({paddleocr_llm_name}): {e}")
+ raise
return None, None, None
diff --git a/test/unit_test/deepdoc/parser/test_mineru_parser.py b/test/unit_test/deepdoc/parser/test_mineru_parser.py
index cc5125033c..5a2d9dddd9 100644
--- a/test/unit_test/deepdoc/parser/test_mineru_parser.py
+++ b/test/unit_test/deepdoc/parser/test_mineru_parser.py
@@ -577,17 +577,22 @@ def test_transfer_to_sections_routes_app_media_separately(monkeypatch, parse_met
class _FakeZipResponse:
"""Stand-in for the streaming response returned by requests.post.
- Provides the minimum surface that _run_mineru_api touches: status code,
- headers (Content-Type), and a `.raw` stream that copyfileobj can drain.
+ Provides the minimum surface that _run_mineru_api touches: status_code,
+ ok, text, headers (Content-Type), and a `.raw` stream that copyfileobj
+ can drain.
"""
- def __init__(self, body: bytes = b"zip-bytes"):
+ def __init__(self, body: bytes = b"zip-bytes", *, status_code: int = 200, text: str = ""):
self._body = body
+ self.status_code = status_code
+ self.ok = 200 <= status_code < 400
+ self.text = text
self.headers = {"Content-Type": "application/zip"}
self.raw = BytesIO(body)
def raise_for_status(self):
- return None
+ if not self.ok:
+ raise RuntimeError(f"HTTP {self.status_code}")
class _FakePostContext:
@@ -602,7 +607,15 @@ class _FakePostContext:
return False
-def _capture_run_mineru_api(monkeypatch, module, *, pdf_path: Path, extracted_dir: Path):
+def _capture_run_mineru_api(
+ monkeypatch,
+ module,
+ *,
+ pdf_path: Path,
+ extracted_dir: Path,
+ status_code: int = 200,
+ text: str = "",
+):
"""Stub everything around requests.post so _run_mineru_api runs end-to-end
against an in-memory response. Returns the captured kwargs dict.
"""
@@ -612,7 +625,7 @@ def _capture_run_mineru_api(monkeypatch, module, *, pdf_path: Path, extracted_di
captured["url"] = url
captured["data"] = data
captured["files"] = files
- return _FakePostContext(_FakeZipResponse(), captured)
+ return _FakePostContext(_FakeZipResponse(status_code=status_code, text=text), captured)
monkeypatch.setattr(module.requests, "post", fake_post)
monkeypatch.setattr(module.os.path, "exists", lambda _p: True)
@@ -720,6 +733,61 @@ def test_end_page_minus_one_normalizes_for_mineru_api(monkeypatch, tmp_path):
assert captured["data"]["end_page_id"] == 12
+def test_run_mineru_api_raises_with_status_and_body_on_non_ok(monkeypatch, tmp_path):
+ module = _load_mineru_parser(monkeypatch)
+ parser = module.MinerUParser(mineru_api="http://mineru.local")
+ parser.mineru_server_url = ""
+
+ pdf_path = tmp_path / "sample.pdf"
+ pdf_path.write_bytes(b"%PDF-1.4 fake")
+ extracted_dir = tmp_path / "out"
+ extracted_dir.mkdir()
+
+ _capture_run_mineru_api(
+ monkeypatch,
+ module,
+ pdf_path=pdf_path,
+ extracted_dir=extracted_dir,
+ status_code=502,
+ text='{"error":"upstream timeout"}',
+ )
+ options = module.MinerUParseOptions()
+
+ with pytest.raises(RuntimeError, match=r"status=502.*upstream timeout") as exc_info:
+ parser._run_mineru_api(pdf_path, extracted_dir, options, callback=None)
+
+ assert "body=" in str(exc_info.value)
+
+
+ ("options_server_url", "parser_server_url", "expected"),
+ [
+ (None, "", None),
+ ("http://options.server", "", "http://options.server"),
+ (None, "http://parser.server", "http://parser.server"),
+ ("http://options.server", "http://parser.server", "http://options.server"),
+ ],
+)
+def test_run_mineru_api_server_url_only_when_set(monkeypatch, tmp_path, options_server_url, parser_server_url, expected):
+ module = _load_mineru_parser(monkeypatch)
+ parser = module.MinerUParser(mineru_api="http://mineru.local", mineru_server_url=parser_server_url)
+
+ pdf_path = tmp_path / "sample.pdf"
+ pdf_path.write_bytes(b"%PDF-1.4 fake")
+ extracted_dir = tmp_path / "out"
+ extracted_dir.mkdir()
+
+ captured = _capture_run_mineru_api(monkeypatch, module, pdf_path=pdf_path, extracted_dir=extracted_dir)
+ options = module.MinerUParseOptions(server_url=options_server_url)
+
+ parser._run_mineru_api(pdf_path, extracted_dir, options, callback=None)
+
+ if expected is None:
+ assert "server_url" not in captured["data"]
+ else:
+ assert captured["data"]["server_url"] == expected
+
+
class _FakePageImage:
def __init__(self, width: int, height: int):
self.size = (width, height)
diff --git a/test/unit_test/rag/test_naive_dispatch.py b/test/unit_test/rag/test_naive_dispatch.py
index b56ce74608..2dd6a2cc05 100644
--- a/test/unit_test/rag/test_naive_dispatch.py
+++ b/test/unit_test/rag/test_naive_dispatch.py
@@ -36,6 +36,7 @@ import importlib.util
import sys
import types
from pathlib import Path
+from unittest.mock import Mock
import pytest
@@ -420,3 +421,37 @@ def test_merge_excel_items_passthrough_when_budget_disabled(naive_module):
items = [("a", (0, 2, 2, 1, 1)), ("b", (0, 3, 3, 1, 1))]
assert naive_module._merge_excel_items(items, chunk_token_num=0) == items
assert naive_module._merge_excel_items([], chunk_token_num=128) == []
+
+
+def test_by_paddleocr_rethrows_parse_failure_and_notifies_callback(naive_module, monkeypatch):
+ callback = Mock()
+ parse_error = RuntimeError("paddle boom")
+ fake_parser = Mock()
+ fake_parser.parse_pdf.side_effect = parse_error
+ fake_bundle = Mock(mdl=fake_parser)
+
+ monkeypatch.setattr(naive_module, "resolve_model_config", Mock(return_value={"id": "cfg"}))
+ monkeypatch.setattr(naive_module, "LLMBundle", Mock(return_value=fake_bundle))
+
+ with pytest.raises(RuntimeError, match="paddle boom"):
+ naive_module.by_paddleocr(
+ "doc.pdf",
+ tenant_id="t1",
+ paddleocr_llm_name="paddle@provider",
+ callback=callback,
+ )
+
+ callback.assert_called_once()
+ prog, msg = callback.call_args.args
+ assert prog == -1
+ assert "Failed to parse pdf via PaddleOCR" in msg
+ assert "paddle boom" in msg
+
+
+def test_by_paddleocr_not_found_path_unchanged(naive_module):
+ callback = Mock()
+
+ sections, tables, parser = naive_module.by_paddleocr("doc.pdf", callback=callback)
+
+ assert (sections, tables, parser) == (None, None, None)
+ callback.assert_called_once_with(-1, "PaddleOCR not found.")