mirror of
https://github.com/paperless-ngx/paperless-ngx.git
synced 2025-09-01 01:46:16 +00:00
Compare commits
7 Commits
fix-joblib
...
fix-sugges
Author | SHA1 | Date | |
---|---|---|---|
![]() |
7ea4893e42 | ||
![]() |
78255d0a99 | ||
![]() |
fc4cb08bda | ||
![]() |
875dc6602b | ||
![]() |
8084ece274 | ||
![]() |
70b24c056b | ||
![]() |
26d2d63c26 |
1008
.github/workflows/ci.yml
vendored
1008
.github/workflows/ci.yml
vendored
File diff suppressed because it is too large
Load Diff
@@ -53,6 +53,7 @@ dependencies = [
|
||||
"ocrmypdf~=16.10.0",
|
||||
"pathvalidate~=3.3.1",
|
||||
"pdf2image~=1.17.0",
|
||||
"psutil>=7",
|
||||
"psycopg-pool",
|
||||
"python-dateutil~=2.9.0",
|
||||
"python-dotenv~=1.1.0",
|
||||
|
@@ -9,8 +9,6 @@ from hashlib import sha256
|
||||
from pathlib import Path
|
||||
from typing import TYPE_CHECKING
|
||||
|
||||
import joblib
|
||||
|
||||
if TYPE_CHECKING:
|
||||
from collections.abc import Iterator
|
||||
from datetime import datetime
|
||||
@@ -53,24 +51,8 @@ class ClassifierModelCorruptError(Exception):
|
||||
pass
|
||||
|
||||
|
||||
def _model_cache_token() -> tuple[str, int, int]:
|
||||
p = Path(settings.MODEL_FILE)
|
||||
if p.exists():
|
||||
try:
|
||||
st = p.stat()
|
||||
return (str(p), int(st.st_mtime), int(st.st_size))
|
||||
except OSError:
|
||||
return (str(p), 0, 0)
|
||||
return (str(p), 0, 0)
|
||||
|
||||
|
||||
@lru_cache(maxsize=1)
|
||||
def _load_classifier_cached(
|
||||
token: tuple[str, int, int],
|
||||
*,
|
||||
raise_exception: bool = False,
|
||||
) -> DocumentClassifier | None:
|
||||
# token used only for cache key; logic depends on current settings
|
||||
def load_classifier(*, raise_exception: bool = False) -> DocumentClassifier | None:
|
||||
if not settings.MODEL_FILE.is_file():
|
||||
logger.debug(
|
||||
"Document classification model does not exist (yet), not "
|
||||
@@ -81,23 +63,25 @@ def _load_classifier_cached(
|
||||
classifier = DocumentClassifier()
|
||||
try:
|
||||
classifier.load()
|
||||
logger.debug("classifier_id=%s", id(classifier))
|
||||
logger.debug(
|
||||
"classifier_data_vectorizer_hash=%s",
|
||||
classifier.data_vectorizer_hash,
|
||||
)
|
||||
|
||||
except IncompatibleClassifierVersionError as e:
|
||||
logger.info(f"Classifier version incompatible: {e.message}, will re-train")
|
||||
try:
|
||||
Path(settings.MODEL_FILE).unlink()
|
||||
except Exception:
|
||||
pass
|
||||
Path(settings.MODEL_FILE).unlink()
|
||||
classifier = None
|
||||
if raise_exception:
|
||||
raise e
|
||||
except ClassifierModelCorruptError as e:
|
||||
# there's something wrong with the model file.
|
||||
logger.exception(
|
||||
"Unrecoverable error while loading document classification model, deleting model file.",
|
||||
"Unrecoverable error while loading document "
|
||||
"classification model, deleting model file.",
|
||||
)
|
||||
try:
|
||||
Path(settings.MODEL_FILE).unlink()
|
||||
except Exception:
|
||||
pass
|
||||
Path(settings.MODEL_FILE).unlink
|
||||
classifier = None
|
||||
if raise_exception:
|
||||
raise e
|
||||
@@ -115,11 +99,6 @@ def _load_classifier_cached(
|
||||
return classifier
|
||||
|
||||
|
||||
def load_classifier(*, raise_exception: bool = False) -> DocumentClassifier | None:
|
||||
token = _model_cache_token()
|
||||
return _load_classifier_cached(token, raise_exception=raise_exception)
|
||||
|
||||
|
||||
class DocumentClassifier:
|
||||
# v7 - Updated scikit-learn package version
|
||||
# v8 - Added storage path classifier
|
||||
@@ -157,48 +136,36 @@ class DocumentClassifier:
|
||||
).hexdigest()
|
||||
|
||||
def load(self) -> None:
|
||||
import joblib
|
||||
from sklearn.exceptions import InconsistentVersionWarning
|
||||
|
||||
# Catch warnings for processing
|
||||
with warnings.catch_warnings(record=True) as w:
|
||||
state = None
|
||||
try:
|
||||
state = joblib.load(settings.MODEL_FILE, mmap_mode="r")
|
||||
except ValueError:
|
||||
# Some environments may fail to mmap small files; fall back to normal load
|
||||
state = joblib.load(settings.MODEL_FILE, mmap_mode=None)
|
||||
except Exception as err:
|
||||
# Fallback to old pickle-based format. Try to read the version and a field to
|
||||
# distinguish truly corrupt files from incompatible versions.
|
||||
# As a fallback, try to detect old pickle-based and mark incompatible
|
||||
try:
|
||||
with Path(settings.MODEL_FILE).open("rb") as f:
|
||||
_version = pickle.load(f)
|
||||
try:
|
||||
_ = pickle.load(f)
|
||||
except Exception as inner:
|
||||
raise ClassifierModelCorruptError from inner
|
||||
# Old, incompatible format
|
||||
raise IncompatibleClassifierVersionError(
|
||||
"Cannot load classifier, incompatible versions.",
|
||||
) from err
|
||||
except (
|
||||
IncompatibleClassifierVersionError,
|
||||
ClassifierModelCorruptError,
|
||||
):
|
||||
_ = pickle.load(f)
|
||||
raise IncompatibleClassifierVersionError(
|
||||
"Cannot load classifier, incompatible versions.",
|
||||
) from err
|
||||
except IncompatibleClassifierVersionError:
|
||||
raise
|
||||
except Exception:
|
||||
# Not even a readable pickle header
|
||||
raise ClassifierModelCorruptError from err
|
||||
|
||||
if (
|
||||
not isinstance(state, dict)
|
||||
or state.get("format_version") != self.FORMAT_VERSION
|
||||
):
|
||||
raise IncompatibleClassifierVersionError(
|
||||
"Cannot load classifier, incompatible versions.",
|
||||
)
|
||||
|
||||
try:
|
||||
if (
|
||||
not isinstance(state, dict)
|
||||
or state.get("format_version") != self.FORMAT_VERSION
|
||||
):
|
||||
raise IncompatibleClassifierVersionError(
|
||||
"Cannot load classifier, incompatible versions.",
|
||||
)
|
||||
|
||||
self.last_doc_change_time = state.get("last_doc_change_time")
|
||||
self.last_auto_type_hash = state.get("last_auto_type_hash")
|
||||
|
||||
@@ -210,6 +177,8 @@ class DocumentClassifier:
|
||||
self.correspondent_classifier = state.get("correspondent_classifier")
|
||||
self.document_type_classifier = state.get("document_type_classifier")
|
||||
self.storage_path_classifier = state.get("storage_path_classifier")
|
||||
except IncompatibleClassifierVersionError:
|
||||
raise
|
||||
except Exception as err:
|
||||
raise ClassifierModelCorruptError from err
|
||||
|
||||
@@ -229,6 +198,8 @@ class DocumentClassifier:
|
||||
raise IncompatibleClassifierVersionError("sklearn version update")
|
||||
|
||||
def save(self) -> None:
|
||||
import joblib
|
||||
|
||||
target_file: Path = settings.MODEL_FILE
|
||||
target_file_temp: Path = target_file.with_suffix(".joblib.part")
|
||||
|
||||
@@ -247,11 +218,6 @@ class DocumentClassifier:
|
||||
joblib.dump(state, target_file_temp, compress=3)
|
||||
|
||||
target_file_temp.rename(target_file)
|
||||
# Invalidate cached classifier loader so subsequent calls see the new file
|
||||
try:
|
||||
_load_classifier_cached.cache_clear()
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
def train(self) -> bool:
|
||||
# Get non-inbox documents
|
||||
|
@@ -370,7 +370,7 @@ class TestClassifier(DirectoriesMixin, TestCase):
|
||||
def test_load_corrupt_file(self, patched_pickle_load: mock.MagicMock):
|
||||
"""
|
||||
GIVEN:
|
||||
- Corrupted legacy classifier pickle file
|
||||
- Corrupted classifier pickle file
|
||||
WHEN:
|
||||
- An attempt is made to load the classifier
|
||||
THEN:
|
||||
@@ -381,10 +381,9 @@ class TestClassifier(DirectoriesMixin, TestCase):
|
||||
# First load is the schema version,allow it
|
||||
patched_pickle_load.side_effect = [DocumentClassifier.FORMAT_VERSION, OSError()]
|
||||
|
||||
# Force the loader down the legacy path by making joblib.load fail
|
||||
with mock.patch("joblib.load", side_effect=Exception("bad joblib")):
|
||||
with self.assertRaises(ClassifierModelCorruptError):
|
||||
self.classifier.load()
|
||||
with self.assertRaises(ClassifierModelCorruptError):
|
||||
self.classifier.load()
|
||||
patched_pickle_load.assert_called()
|
||||
|
||||
patched_pickle_load.reset_mock()
|
||||
patched_pickle_load.side_effect = [
|
||||
@@ -392,8 +391,8 @@ class TestClassifier(DirectoriesMixin, TestCase):
|
||||
ClassifierModelCorruptError(),
|
||||
]
|
||||
|
||||
with mock.patch("joblib.load", side_effect=Exception("bad joblib")):
|
||||
self.assertIsNone(load_classifier())
|
||||
self.assertIsNone(load_classifier())
|
||||
patched_pickle_load.assert_called()
|
||||
|
||||
def test_load_new_scikit_learn_version(self):
|
||||
"""
|
||||
|
@@ -3,7 +3,9 @@ import logging
|
||||
import os
|
||||
import platform
|
||||
import re
|
||||
import resource
|
||||
import tempfile
|
||||
import time
|
||||
import zipfile
|
||||
from datetime import datetime
|
||||
from pathlib import Path
|
||||
@@ -190,6 +192,33 @@ if settings.AUDIT_LOG_ENABLED:
|
||||
|
||||
logger = logging.getLogger("paperless.api")
|
||||
|
||||
try:
|
||||
import psutil
|
||||
|
||||
_PS = psutil.Process(os.getpid())
|
||||
except Exception:
|
||||
_PS = None
|
||||
|
||||
_diag_log = logging.getLogger("paperless")
|
||||
|
||||
|
||||
def _mem_mb():
|
||||
rss = _PS.memory_info().rss if _PS else 0
|
||||
peak_kb = resource.getrusage(resource.RUSAGE_SELF).ru_maxrss
|
||||
return rss / (1024 * 1024), peak_kb / 1024.0
|
||||
|
||||
|
||||
def _mark(phase, doc_id, t0):
|
||||
rss, peak = _mem_mb()
|
||||
_diag_log.debug(
|
||||
"sugg doc=%s phase=%s rss=%.1fMB peak=%.1fMB t=%.1fms",
|
||||
doc_id,
|
||||
phase,
|
||||
rss,
|
||||
peak,
|
||||
(time.perf_counter() - t0) * 1000,
|
||||
)
|
||||
|
||||
|
||||
class IndexView(TemplateView):
|
||||
template_name = "index.html"
|
||||
@@ -758,7 +787,16 @@ class DocumentViewSet(
|
||||
),
|
||||
)
|
||||
def suggestions(self, request, pk=None):
|
||||
doc = get_object_or_404(Document.objects.select_related("owner"), pk=pk)
|
||||
t0 = time.perf_counter()
|
||||
# Don't fetch content here
|
||||
doc = get_object_or_404(
|
||||
Document.objects.select_related("owner").only(
|
||||
"id",
|
||||
"owner_id",
|
||||
),
|
||||
pk=pk,
|
||||
)
|
||||
_mark("start", doc.pk, t0)
|
||||
if request.user is not None and not has_perms_owner_aware(
|
||||
request.user,
|
||||
"view_document",
|
||||
@@ -769,18 +807,23 @@ class DocumentViewSet(
|
||||
document_suggestions = get_suggestion_cache(doc.pk)
|
||||
|
||||
if document_suggestions is not None:
|
||||
_mark("cache_hit_return", doc.pk, t0)
|
||||
refresh_suggestions_cache(doc.pk)
|
||||
return Response(document_suggestions.suggestions)
|
||||
|
||||
classifier = load_classifier()
|
||||
_mark("loaded_classifier", doc.pk, t0)
|
||||
|
||||
dates = []
|
||||
if settings.NUMBER_OF_SUGGESTED_DATES > 0:
|
||||
gen = parse_date_generator(doc.filename, doc.content)
|
||||
_mark("before_dates", doc.pk, t0)
|
||||
dates = sorted(
|
||||
{i for i in itertools.islice(gen, settings.NUMBER_OF_SUGGESTED_DATES)},
|
||||
)
|
||||
_mark("after_dates", doc.pk, t0)
|
||||
|
||||
_mark("before_match", doc.pk, t0)
|
||||
resp_data = {
|
||||
"correspondents": [
|
||||
c.id for c in match_correspondents(doc, classifier, request.user)
|
||||
@@ -794,9 +837,11 @@ class DocumentViewSet(
|
||||
],
|
||||
"dates": [date.strftime("%Y-%m-%d") for date in dates if date is not None],
|
||||
}
|
||||
_mark("assembled_resp", doc.pk, t0)
|
||||
|
||||
# Cache the suggestions and the classifier hash for later
|
||||
set_suggestions_cache(doc.pk, resp_data, classifier)
|
||||
_mark("cached", doc.pk, t0)
|
||||
|
||||
return Response(resp_data)
|
||||
|
||||
|
@@ -1,7 +1,14 @@
|
||||
import logging
|
||||
import os
|
||||
import resource
|
||||
import time
|
||||
|
||||
from django.conf import settings
|
||||
|
||||
from paperless import version
|
||||
|
||||
logger = logging.getLogger("middleware")
|
||||
|
||||
|
||||
class ApiVersionMiddleware:
|
||||
def __init__(self, get_response):
|
||||
@@ -15,3 +22,56 @@ class ApiVersionMiddleware:
|
||||
response["X-Version"] = version.__full_version_str__
|
||||
|
||||
return response
|
||||
|
||||
|
||||
try:
|
||||
import psutil
|
||||
|
||||
_PSUTIL = True
|
||||
except Exception:
|
||||
_PSUTIL = False
|
||||
|
||||
|
||||
class MemLogMiddleware:
|
||||
def __init__(self, get_response):
|
||||
self.get_response = get_response
|
||||
|
||||
def __call__(self, request):
|
||||
# capture baseline
|
||||
ru_before = resource.getrusage(resource.RUSAGE_SELF).ru_maxrss
|
||||
if _PSUTIL:
|
||||
p = psutil.Process()
|
||||
rss_before = p.memory_info().rss
|
||||
else:
|
||||
rss_before = 0
|
||||
|
||||
t0 = time.perf_counter()
|
||||
try:
|
||||
return self.get_response(request)
|
||||
finally:
|
||||
dur_ms = (time.perf_counter() - t0) * 1000.0
|
||||
|
||||
ru_after = resource.getrusage(resource.RUSAGE_SELF).ru_maxrss
|
||||
# ru_maxrss is KB on Linux; convert to MB
|
||||
peak_mb = (ru_after) / 1024.0
|
||||
peak_delta_mb = (ru_after - ru_before) / 1024.0
|
||||
|
||||
if _PSUTIL:
|
||||
rss_after = p.memory_info().rss
|
||||
delta_mb = (rss_after - rss_before) / (1024 * 1024)
|
||||
rss_mb = rss_after / (1024 * 1024)
|
||||
else:
|
||||
delta_mb = 0.0
|
||||
rss_mb = 0.0
|
||||
|
||||
logger.debug(
|
||||
"pid=%s mem rss=%.1fMB Δend=%.1fMB peak=%.1fMB Δpeak=%.1fMB dur=%.1fms %s %s",
|
||||
os.getpid(),
|
||||
rss_mb,
|
||||
delta_mb,
|
||||
peak_mb,
|
||||
peak_delta_mb,
|
||||
dur_ms,
|
||||
request.method,
|
||||
request.path,
|
||||
)
|
||||
|
@@ -363,6 +363,7 @@ if DEBUG:
|
||||
)
|
||||
|
||||
MIDDLEWARE = [
|
||||
"paperless.middleware.MemLogMiddleware",
|
||||
"django.middleware.security.SecurityMiddleware",
|
||||
"whitenoise.middleware.WhiteNoiseMiddleware",
|
||||
"django.contrib.sessions.middleware.SessionMiddleware",
|
||||
@@ -833,7 +834,7 @@ LOGGING = {
|
||||
"disable_existing_loggers": False,
|
||||
"formatters": {
|
||||
"verbose": {
|
||||
"format": "[{asctime}] [{levelname}] [{name}] {message}",
|
||||
"format": "[{asctime}] [{levelname}] [{name}] pid={process} {message}",
|
||||
"style": "{",
|
||||
},
|
||||
"simple": {
|
||||
@@ -878,6 +879,7 @@ LOGGING = {
|
||||
"kombu": {"handlers": ["file_celery"], "level": "DEBUG"},
|
||||
"_granian": {"handlers": ["file_paperless"], "level": "DEBUG"},
|
||||
"granian.access": {"handlers": ["file_paperless"], "level": "DEBUG"},
|
||||
"middleware": {"handlers": ["console"], "level": "DEBUG"},
|
||||
},
|
||||
}
|
||||
|
||||
|
15
uv.lock
generated
15
uv.lock
generated
@@ -2046,6 +2046,7 @@ dependencies = [
|
||||
{ name = "ocrmypdf", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "pathvalidate", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "pdf2image", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "psutil", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "psycopg-pool", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "python-dateutil", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
{ name = "python-dotenv", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
|
||||
@@ -2182,6 +2183,7 @@ requires-dist = [
|
||||
{ name = "ocrmypdf", specifier = "~=16.10.0" },
|
||||
{ name = "pathvalidate", specifier = "~=3.3.1" },
|
||||
{ name = "pdf2image", specifier = "~=1.17.0" },
|
||||
{ name = "psutil", specifier = ">=7.0.0" },
|
||||
{ name = "psycopg", extras = ["c", "pool"], marker = "extra == 'postgres'", specifier = "==3.2.9" },
|
||||
{ name = "psycopg-c", marker = "python_full_version == '3.12.*' and platform_machine == 'aarch64' and sys_platform == 'linux' and extra == 'postgres'", url = "https://github.com/paperless-ngx/builder/releases/download/psycopg-3.2.9/psycopg_c-3.2.9-cp312-cp312-linux_aarch64.whl" },
|
||||
{ name = "psycopg-c", marker = "python_full_version == '3.12.*' and platform_machine == 'x86_64' and sys_platform == 'linux' and extra == 'postgres'", url = "https://github.com/paperless-ngx/builder/releases/download/psycopg-3.2.9/psycopg_c-3.2.9-cp312-cp312-linux_x86_64.whl" },
|
||||
@@ -2548,6 +2550,19 @@ wheels = [
|
||||
{ url = "https://files.pythonhosted.org/packages/e4/ea/d836f008d33151c7a1f62caf3d8dd782e4d15f6a43897f64480c2b8de2ad/prompt_toolkit-3.0.50-py3-none-any.whl", hash = "sha256:9b6427eb19e479d98acff65196a307c555eb567989e6d88ebbb1b509d9779198", size = 387816, upload-time = "2025-01-20T15:55:29.98Z" },
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "psutil"
|
||||
version = "7.0.0"
|
||||
source = { registry = "https://pypi.org/simple" }
|
||||
sdist = { url = "https://files.pythonhosted.org/packages/2a/80/336820c1ad9286a4ded7e845b2eccfcb27851ab8ac6abece774a6ff4d3de/psutil-7.0.0.tar.gz", hash = "sha256:7be9c3eba38beccb6495ea33afd982a44074b78f28c434a1f51cc07fd315c456", size = 497003, upload-time = "2025-02-13T21:54:07.946Z" }
|
||||
wheels = [
|
||||
{ url = "https://files.pythonhosted.org/packages/ed/e6/2d26234410f8b8abdbf891c9da62bee396583f713fb9f3325a4760875d22/psutil-7.0.0-cp36-abi3-macosx_10_9_x86_64.whl", hash = "sha256:101d71dc322e3cffd7cea0650b09b3d08b8e7c4109dd6809fe452dfd00e58b25", size = 238051, upload-time = "2025-02-13T21:54:12.36Z" },
|
||||
{ url = "https://files.pythonhosted.org/packages/04/8b/30f930733afe425e3cbfc0e1468a30a18942350c1a8816acfade80c005c4/psutil-7.0.0-cp36-abi3-macosx_11_0_arm64.whl", hash = "sha256:39db632f6bb862eeccf56660871433e111b6ea58f2caea825571951d4b6aa3da", size = 239535, upload-time = "2025-02-13T21:54:16.07Z" },
|
||||
{ url = "https://files.pythonhosted.org/packages/2a/ed/d362e84620dd22876b55389248e522338ed1bf134a5edd3b8231d7207f6d/psutil-7.0.0-cp36-abi3-manylinux_2_12_i686.manylinux2010_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:1fcee592b4c6f146991ca55919ea3d1f8926497a713ed7faaf8225e174581e91", size = 275004, upload-time = "2025-02-13T21:54:18.662Z" },
|
||||
{ url = "https://files.pythonhosted.org/packages/bf/b9/b0eb3f3cbcb734d930fdf839431606844a825b23eaf9a6ab371edac8162c/psutil-7.0.0-cp36-abi3-manylinux_2_12_x86_64.manylinux2010_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:4b1388a4f6875d7e2aff5c4ca1cc16c545ed41dd8bb596cefea80111db353a34", size = 277986, upload-time = "2025-02-13T21:54:21.811Z" },
|
||||
{ url = "https://files.pythonhosted.org/packages/eb/a2/709e0fe2f093556c17fbafda93ac032257242cabcc7ff3369e2cb76a97aa/psutil-7.0.0-cp36-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:a5f098451abc2828f7dc6b58d44b532b22f2088f4999a937557b603ce72b1993", size = 279544, upload-time = "2025-02-13T21:54:24.68Z" },
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "psycopg"
|
||||
version = "3.2.9"
|
||||
|
Reference in New Issue
Block a user