This commit is contained in:
Jens
+1
-1
@@ -7,7 +7,7 @@ DJANGO_ALLOWED_HOSTS=localhost,127.0.0.1
|
||||
DJANGO_CSRF_TRUSTED_ORIGINS=http://localhost:1226
|
||||
DJANGO_TIME_ZONE=Europe/Brussels
|
||||
VACATURERADAR_OWNER_NAME=Jens
|
||||
VACATURERADAR_VERSION=0.3.13
|
||||
VACATURERADAR_VERSION=0.3.14
|
||||
|
||||
# Gebruik DATABASE_URL niet voor SQLite. Laat leeg voor lokale sqlite-ontwikkeling.
|
||||
POSTGRES_DB=vacatureradar
|
||||
|
||||
@@ -2,6 +2,18 @@
|
||||
|
||||
Alle betekenisvolle wijzigingen worden hier bijgehouden. Het project volgt voorlopig een pre-1.0 semantische versieaanpak.
|
||||
|
||||
## 0.3.14 - 2026-07-29
|
||||
|
||||
### Gewijzigd
|
||||
|
||||
- De importpipeline hergebruikt werkgevers en actieve profielen binnen één run, synchroniseert provenance gebundeld en vermijdt volledige rescoring van inhoudelijk ongewijzigde vacatures terwijl ScoreRun-historiek behouden blijft.
|
||||
- `last_changed` wijzigt alleen nog bij een inhoudelijke vacaturewijziging; een nieuwe waarneming actualiseert uitsluitend `last_seen`.
|
||||
- Een fail-closed PostgreSQL-runner levert één warm-up, drie metingen, p50/p95, throughput, geheugen en queryfamilies als machineleesbaar rapport.
|
||||
|
||||
### Performance
|
||||
|
||||
- De officiële 250-itemmeting is gemiddeld 44,74% sneller, gebruikt minstens 63,73% minder queries en verwerkt 83,42% meer documenten per seconde dan 0.3.13.
|
||||
|
||||
## Niet uitgebracht
|
||||
|
||||
### Toegevoegd
|
||||
|
||||
+2
-2
@@ -1,5 +1,5 @@
|
||||
# Huidige toestand
|
||||
|
||||
VacatureRadar 0.3.13 is een single-user-first, server-rendered vacature-intelligencecockpit. De actuele uitvoeringsstatus, laatste gate en externe restpunten staan in [docs/ai/PROJECT_STATE.md](docs/ai/PROJECT_STATE.md). Productgebruik en installatie beginnen in [README.md](README.md).
|
||||
VacatureRadar 0.3.14 is een single-user-first, server-rendered vacature-intelligencecockpit. De actuele uitvoeringsstatus, laatste gate en externe restpunten staan in [docs/ai/PROJECT_STATE.md](docs/ai/PROJECT_STATE.md). Productgebruik en installatie beginnen in [README.md](README.md).
|
||||
|
||||
Versie 0.3.13 draait publiek op `https://vacatureradar.itworx.tech` als Unraid AIO-image `sha256:34aef8853f77a3f311b119bd67d9f1160a97fd4a21e62c9776222be289bf3b88`. TLS, proxy, secure cookies, migrations, containerrestart, kernflows, drie responsive viewports, back-up/restore en rollback zijn op 2026-07-29 live bewezen. Alleen optionele externe identity- en mailboxcredentials staan nog in [USER_INPUT_REQUIRED.md](USER_INPUT_REQUIRED.md).
|
||||
Release 0.3.14 verlaagt voor de vaste 250-itemimport de gemiddelde duur met 44,74% en het queryvolume met minstens 63,73%, met behoud van parser-, dedupe-, ranking-, provenance- en scorehistoriekcontracten. Het actuele productie-image en livebewijs worden na deployment in [docs/ai/PROJECT_STATE.md](docs/ai/PROJECT_STATE.md) geregistreerd. Alleen optionele externe identity- en mailboxcredentials staan nog in [USER_INPUT_REQUIRED.md](USER_INPUT_REQUIRED.md).
|
||||
|
||||
+148
-23
@@ -1,12 +1,21 @@
|
||||
from __future__ import annotations
|
||||
|
||||
from dataclasses import dataclass
|
||||
from dataclasses import field as dataclass_field
|
||||
from decimal import Decimal
|
||||
from urllib.parse import urlsplit
|
||||
|
||||
from django.db import IntegrityError, transaction
|
||||
from django.utils import timezone
|
||||
|
||||
from apps.jobs.models import Employer, FieldProvenance, JobPosting, JobSourceAlias, JobVersion
|
||||
from apps.jobs.models import (
|
||||
Employer,
|
||||
FieldProvenance,
|
||||
JobPosting,
|
||||
JobSourceAlias,
|
||||
JobVersion,
|
||||
ScoreRun,
|
||||
)
|
||||
from apps.profiles.models import SearchProfile
|
||||
from apps.sources.adapters.base import FieldEvidence
|
||||
from apps.sources.adapters.registry import registry
|
||||
@@ -29,17 +38,31 @@ RECRUITER_TERMS = {
|
||||
}
|
||||
|
||||
|
||||
@dataclass
|
||||
class PersistenceContext:
|
||||
"""Run-scoped reference cache; never shared between workers or imports."""
|
||||
|
||||
employers: dict[tuple[str, str], Employer | None] = dataclass_field(default_factory=dict)
|
||||
active_profiles: list[SearchProfile] | None = None
|
||||
latest_scores: dict[tuple[object, int], ScoreRun] = dataclass_field(default_factory=dict)
|
||||
|
||||
|
||||
def _confidence(value: float) -> Decimal:
|
||||
return Decimal(str(max(0.0, min(1.0, value))))
|
||||
|
||||
|
||||
def resolve_employer(draft: CanonicalJobDraft) -> Employer | None:
|
||||
def resolve_employer(
|
||||
draft: CanonicalJobDraft, *, context: PersistenceContext | None = None
|
||||
) -> Employer | None:
|
||||
name = draft.employer_name.strip()
|
||||
domain = draft.employer_domain.strip()
|
||||
if not name and not domain:
|
||||
return None
|
||||
display_name = name or domain
|
||||
normalized = normalize_token(display_name)
|
||||
cache_key = (normalized, domain)
|
||||
if context is not None and cache_key in context.employers:
|
||||
return context.employers[cache_key]
|
||||
recruiter = any(term in normalized for term in RECRUITER_TERMS)
|
||||
employer, _ = Employer.objects.get_or_create(
|
||||
normalized_name=normalized,
|
||||
@@ -61,6 +84,8 @@ def resolve_employer(draft: CanonicalJobDraft) -> Employer | None:
|
||||
changed.extend(["is_recruiter", "is_direct_employer"])
|
||||
if changed:
|
||||
employer.save(update_fields=[*changed, "updated_at"])
|
||||
if context is not None:
|
||||
context.employers[cache_key] = employer
|
||||
return employer
|
||||
|
||||
|
||||
@@ -162,12 +187,81 @@ def _apply_draft(
|
||||
if direct and canonical_url and job.canonical_url != canonical_url:
|
||||
job.canonical_url = canonical_url
|
||||
changed.append("canonical_url")
|
||||
if changed and "last_changed" not in changed:
|
||||
substantive_changes = [field for field in changed if field != "last_seen"]
|
||||
if substantive_changes and "last_changed" not in changed:
|
||||
job.last_changed = timezone.now()
|
||||
changed.append("last_changed")
|
||||
return changed
|
||||
|
||||
|
||||
def _sync_evidence(
|
||||
*,
|
||||
job: JobPosting,
|
||||
alias: JobSourceAlias,
|
||||
evidence_items: list[FieldEvidence],
|
||||
parser_version: str,
|
||||
) -> None:
|
||||
existing = {
|
||||
(item.field_name, item.extraction_method): item
|
||||
for item in FieldProvenance.objects.filter(job=job, source_alias=alias)
|
||||
}
|
||||
creates: list[FieldProvenance] = []
|
||||
updates: list[FieldProvenance] = []
|
||||
for evidence in evidence_items:
|
||||
key = (evidence.field_name, evidence.method)
|
||||
values = {
|
||||
"confidence": _confidence(evidence.confidence),
|
||||
"evidence_excerpt": evidence.evidence[:1000],
|
||||
"parser_version": parser_version,
|
||||
}
|
||||
current = existing.get(key)
|
||||
if current is None:
|
||||
creates.append(
|
||||
FieldProvenance(
|
||||
job=job,
|
||||
source_alias=alias,
|
||||
field_name=evidence.field_name,
|
||||
extraction_method=evidence.method,
|
||||
**values,
|
||||
)
|
||||
)
|
||||
continue
|
||||
changed = False
|
||||
for field_name, value in values.items():
|
||||
if getattr(current, field_name) != value:
|
||||
setattr(current, field_name, value)
|
||||
changed = True
|
||||
if changed:
|
||||
current.updated_at = timezone.now()
|
||||
updates.append(current)
|
||||
if creates:
|
||||
FieldProvenance.objects.bulk_create(creates, batch_size=250)
|
||||
if updates:
|
||||
FieldProvenance.objects.bulk_update(
|
||||
updates,
|
||||
["confidence", "evidence_excerpt", "parser_version", "updated_at"],
|
||||
batch_size=250,
|
||||
)
|
||||
|
||||
|
||||
def _copy_score(score: ScoreRun) -> ScoreRun:
|
||||
return ScoreRun.objects.create(
|
||||
job=score.job,
|
||||
profile=score.profile,
|
||||
profile_version=score.profile_version,
|
||||
score=score.score,
|
||||
confidence=score.confidence,
|
||||
recommendation=score.recommendation,
|
||||
components=score.components,
|
||||
positives=score.positives,
|
||||
concerns=score.concerns,
|
||||
hard_exclusions=score.hard_exclusions,
|
||||
evidence=score.evidence,
|
||||
model_version=score.model_version,
|
||||
prompt_version=score.prompt_version,
|
||||
)
|
||||
|
||||
|
||||
@transaction.atomic
|
||||
def persist_draft(
|
||||
draft: CanonicalJobDraft,
|
||||
@@ -176,14 +270,16 @@ def persist_draft(
|
||||
parser_key: str,
|
||||
parser_version: str,
|
||||
extraction_confidence: float,
|
||||
context: PersistenceContext | None = None,
|
||||
) -> tuple[JobPosting, DedupeDecision, bool]:
|
||||
source = document.source
|
||||
employer = resolve_employer(draft)
|
||||
employer = resolve_employer(draft, context=context)
|
||||
decision = find_existing_job(draft, source=source)
|
||||
direct = _source_is_direct(source, draft)
|
||||
if decision.resolved_direct:
|
||||
direct = True
|
||||
created = False
|
||||
substantive_change = False
|
||||
|
||||
if decision.job is None:
|
||||
try:
|
||||
@@ -240,6 +336,7 @@ def persist_draft(
|
||||
if extraction_confidence > float(job.extraction_confidence):
|
||||
job.extraction_confidence = _confidence(extraction_confidence)
|
||||
changed.append("extraction_confidence")
|
||||
substantive_change = any(field not in {"last_seen", "last_changed"} for field in changed)
|
||||
if changed:
|
||||
job.save(update_fields=list(dict.fromkeys([*changed, "updated_at"])))
|
||||
|
||||
@@ -284,31 +381,58 @@ def persist_draft(
|
||||
update_fields=["last_seen", "raw_document", "payload", "is_canonical", "updated_at"]
|
||||
)
|
||||
|
||||
for evidence in draft.evidence:
|
||||
FieldProvenance.objects.update_or_create(
|
||||
job=job,
|
||||
source_alias=alias,
|
||||
field_name=evidence.field_name,
|
||||
extraction_method=evidence.method,
|
||||
defaults={
|
||||
"confidence": _confidence(evidence.confidence),
|
||||
"evidence_excerpt": evidence.evidence[:1000],
|
||||
"parser_version": parser_version,
|
||||
},
|
||||
)
|
||||
|
||||
JobVersion.objects.get_or_create(
|
||||
_sync_evidence(
|
||||
job=job,
|
||||
content_hash=job.content_hash,
|
||||
defaults={"snapshot": job_snapshot(job), "changed_fields": []},
|
||||
alias=alias,
|
||||
evidence_items=draft.evidence,
|
||||
parser_version=parser_version,
|
||||
)
|
||||
for profile in SearchProfile.objects.filter(is_active=True):
|
||||
score_and_save(job, profile)
|
||||
|
||||
if created or substantive_change:
|
||||
JobVersion.objects.get_or_create(
|
||||
job=job,
|
||||
content_hash=job.content_hash,
|
||||
defaults={"snapshot": job_snapshot(job), "changed_fields": []},
|
||||
)
|
||||
if context is not None:
|
||||
if context.active_profiles is None:
|
||||
context.active_profiles = list(SearchProfile.objects.filter(is_active=True))
|
||||
profiles = context.active_profiles
|
||||
else:
|
||||
profiles = SearchProfile.objects.filter(is_active=True)
|
||||
for profile in profiles:
|
||||
score = score_and_save(job, profile)
|
||||
if context is not None:
|
||||
context.latest_scores[(job.pk, profile.pk)] = score
|
||||
else:
|
||||
if context is not None:
|
||||
if context.active_profiles is None:
|
||||
context.active_profiles = list(SearchProfile.objects.filter(is_active=True))
|
||||
profiles = context.active_profiles
|
||||
else:
|
||||
profiles = SearchProfile.objects.filter(is_active=True)
|
||||
for profile in profiles:
|
||||
cache_key = (job.pk, profile.pk)
|
||||
previous = context.latest_scores.get(cache_key) if context is not None else None
|
||||
if previous is None:
|
||||
previous = (
|
||||
ScoreRun.objects.filter(job=job, profile=profile)
|
||||
.order_by("-created_at")
|
||||
.first()
|
||||
)
|
||||
if previous is None or previous.profile_version != profile.version:
|
||||
score = score_and_save(job, profile)
|
||||
else:
|
||||
score = _copy_score(previous)
|
||||
if context is not None:
|
||||
context.latest_scores[cache_key] = score
|
||||
return job, decision, created
|
||||
|
||||
|
||||
@transaction.atomic
|
||||
def process_raw_document(document: RawDocument) -> dict[str, int | str | list[str]]:
|
||||
def process_raw_document(
|
||||
document: RawDocument, *, context: PersistenceContext | None = None
|
||||
) -> dict[str, int | str | list[str]]:
|
||||
result = registry.extract(document)
|
||||
document.parser_key = result.parser_key
|
||||
document.parser_version = result.parser_version
|
||||
@@ -341,6 +465,7 @@ def process_raw_document(document: RawDocument) -> dict[str, int | str | list[st
|
||||
parser_key=result.parser_key,
|
||||
parser_version=result.parser_version,
|
||||
extraction_confidence=result.confidence,
|
||||
context=context,
|
||||
)
|
||||
if was_created:
|
||||
created += 1
|
||||
|
||||
@@ -0,0 +1,745 @@
|
||||
{
|
||||
"configuration": {
|
||||
"concurrency": 1,
|
||||
"dataset": "fixtures/benchmark/quality_benchmark.json",
|
||||
"iterations": 3,
|
||||
"jobs": 250,
|
||||
"warmup_runs": 1
|
||||
},
|
||||
"database_identity": {
|
||||
"engine": "django.db.backends.postgresql",
|
||||
"host": "127.0.0.1",
|
||||
"name_prefix": "vacatureradar_bench_"
|
||||
},
|
||||
"iterations": [
|
||||
{
|
||||
"dataset_version": "vr116-2026-07-21",
|
||||
"dedupe": {
|
||||
"case_reports": [
|
||||
{
|
||||
"decision": "exact_external_id",
|
||||
"decision_similarity": 1.0,
|
||||
"expected_match": "seed-infra",
|
||||
"expected_match_present": true,
|
||||
"id": "D-001",
|
||||
"is_expected_match": true,
|
||||
"matched": true,
|
||||
"matched_job_id": "267f0fd5-4543-404c-acd1-c6f6ee72e4c0",
|
||||
"query": "D-001"
|
||||
},
|
||||
{
|
||||
"decision": "exact_external_id",
|
||||
"decision_similarity": 1.0,
|
||||
"expected_match": "seed-network",
|
||||
"expected_match_present": true,
|
||||
"id": "D-002",
|
||||
"is_expected_match": true,
|
||||
"matched": true,
|
||||
"matched_job_id": "6a196e1a-bed6-44d0-94df-c2578c367673",
|
||||
"query": "D-002"
|
||||
},
|
||||
{
|
||||
"decision": "new",
|
||||
"decision_similarity": 0.0,
|
||||
"expected_match": null,
|
||||
"expected_match_present": false,
|
||||
"id": "D-003",
|
||||
"is_expected_match": false,
|
||||
"matched": false,
|
||||
"matched_job_id": null,
|
||||
"query": "D-003"
|
||||
},
|
||||
{
|
||||
"decision": "new",
|
||||
"decision_similarity": 0.0,
|
||||
"expected_match": null,
|
||||
"expected_match_present": false,
|
||||
"id": "D-004",
|
||||
"is_expected_match": false,
|
||||
"matched": false,
|
||||
"matched_job_id": null,
|
||||
"query": "D-004"
|
||||
}
|
||||
],
|
||||
"false_merges": 0,
|
||||
"false_negative": 0,
|
||||
"false_positive": 0,
|
||||
"missed_merges": 0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"seed_jobs": 2,
|
||||
"status": "passed",
|
||||
"true_negative": 2,
|
||||
"true_positive": 2
|
||||
},
|
||||
"hardware": {
|
||||
"cpu_count": 20,
|
||||
"debug_mode": true,
|
||||
"platform": "Linux-6.12.54-Unraid-x86_64-with-glibc2.41",
|
||||
"python": "3.13.14",
|
||||
"release": "6.12.54-Unraid"
|
||||
},
|
||||
"jobs_processed": 250,
|
||||
"nfr_007": {
|
||||
"measured": false,
|
||||
"p95_ms": null,
|
||||
"passed": false
|
||||
},
|
||||
"parser": {
|
||||
"cases": [
|
||||
{
|
||||
"coverage_ratio": 1.0,
|
||||
"covered_fields": 6,
|
||||
"expected_fields": 6,
|
||||
"expected_job_count": 1,
|
||||
"expected_parser_key": "jsonld-jobposting",
|
||||
"expected_parser_version": "1.0.0",
|
||||
"extracted": 1,
|
||||
"fixture": "fixtures/pages/sample_jsonld_job.html",
|
||||
"id": "P-001",
|
||||
"parser_key": "jsonld-jobposting",
|
||||
"parser_version": "1.0.0",
|
||||
"passed": true,
|
||||
"warnings": []
|
||||
},
|
||||
{
|
||||
"coverage_ratio": 1.0,
|
||||
"covered_fields": 6,
|
||||
"expected_fields": 6,
|
||||
"expected_job_count": 1,
|
||||
"expected_parser_key": "generic-html",
|
||||
"expected_parser_version": "1.1.0",
|
||||
"extracted": 1,
|
||||
"fixture": "fixtures/pages/sample_generic_job.html",
|
||||
"id": "P-002",
|
||||
"parser_key": "generic-html",
|
||||
"parser_version": "1.1.0",
|
||||
"passed": true,
|
||||
"warnings": []
|
||||
}
|
||||
],
|
||||
"coverage_ratio": 1.0,
|
||||
"covered_fields": 12,
|
||||
"expected_fields": 12,
|
||||
"status": "passed",
|
||||
"unknown_data_ratio": 0.0
|
||||
},
|
||||
"performance": {
|
||||
"avg_scores_detail_ms": 50.936,
|
||||
"avg_scores_import_ms": 30.859,
|
||||
"avg_scores_list_ms": 104.346,
|
||||
"avg_scores_rescore_ms": 14.893,
|
||||
"dashboard": {
|
||||
"detail_p95_ms": 55.28,
|
||||
"detail_query_count": 25,
|
||||
"detail_query_types": {
|
||||
"SELECT": 25
|
||||
},
|
||||
"list_p95_ms": 135.076,
|
||||
"list_query_count": 76,
|
||||
"list_query_types": {
|
||||
"SELECT": 76
|
||||
},
|
||||
"timing_count": 50
|
||||
},
|
||||
"import": {
|
||||
"duration_ms_total": 7714.791,
|
||||
"p50_ms": 29.722,
|
||||
"p95_ms": 39.37,
|
||||
"query_count": 3264,
|
||||
"query_types": {
|
||||
"INSERT": 1003,
|
||||
"RELEASE": 253,
|
||||
"SAVEPOINT": 253,
|
||||
"SELECT": 1256,
|
||||
"UPDATE": 499
|
||||
},
|
||||
"timing_count": 250
|
||||
},
|
||||
"jobs": 1,
|
||||
"performance": {
|
||||
"memory_current_mb": 12.418,
|
||||
"memory_peak_mb": 17.349,
|
||||
"nfr_007": true,
|
||||
"p50_ms": 30.306,
|
||||
"p95_ms": 94.573,
|
||||
"total_duration_ms": 11611.716
|
||||
},
|
||||
"query_count": 3370,
|
||||
"rescore": {
|
||||
"duration_ms_total": 14.893,
|
||||
"p50_ms": 14.893,
|
||||
"p95_ms": 14.893,
|
||||
"query_count": 5,
|
||||
"query_types": {
|
||||
"INSERT": 1,
|
||||
"SELECT": 4
|
||||
},
|
||||
"timing_count": 1
|
||||
},
|
||||
"source_run_id": 1,
|
||||
"top_n": 25
|
||||
},
|
||||
"ranking": {
|
||||
"churn_ratio": 0.0,
|
||||
"post_top_n": [
|
||||
"8f49a32b-f425-425e-ac47-fbd1e1ab020f",
|
||||
"8302cd44-569c-4109-83dd-87663d053200",
|
||||
"fb63a999-5aa0-4912-8d06-400f852676cc",
|
||||
"1476c46b-84ce-4151-b777-9dac8b533a8c",
|
||||
"518e4b70-6742-43d2-bd39-90b3af026d89",
|
||||
"9fa543b2-d5c9-47e1-8cba-ed1e1bfbfb86",
|
||||
"5759f1c8-4974-4333-98c4-2ab0aec97918",
|
||||
"77565fbe-27ea-43ee-bcb7-a218b073af9f",
|
||||
"a9f67c53-f22d-4d72-be23-2ef95eb49443",
|
||||
"ff866750-238b-4eb6-a9b7-7cacd8c6b9d0",
|
||||
"8470bf3b-1760-4d4c-8755-387fcc5c92e8",
|
||||
"293b9d4a-be83-4c72-b08e-02d6170ac405",
|
||||
"d00b4e66-dfd1-45e8-af78-b643f59b5327",
|
||||
"de6cbb4d-5225-4975-91f3-23b280492598",
|
||||
"bb5bfca7-38d3-4639-83bd-da639ce2b4b5",
|
||||
"d9a6079c-eb37-47e0-b588-28f87ebc7411",
|
||||
"2d83e2f0-9a87-40a7-89cc-2f0689f52f18",
|
||||
"e985ad1c-5bd1-4a29-b683-f1be109998ba",
|
||||
"081bdfa8-0dfc-43d1-9b85-44cfe72892c1",
|
||||
"79b5af36-6f76-4e35-af5c-ed3ca755b3ac",
|
||||
"9c7d00f4-d307-42f3-b6d4-f95b2a6f7597",
|
||||
"1053c44f-6f09-4a1e-b43f-f858fcb2441c",
|
||||
"5f5005e9-6fe6-435a-992f-46440a803197",
|
||||
"a9fe6c71-e1e1-423f-9811-5318c0b49688",
|
||||
"f06e60ed-6d6d-44ef-b2f9-3ef57dc94cdf"
|
||||
],
|
||||
"pre_top_n": [
|
||||
"8f49a32b-f425-425e-ac47-fbd1e1ab020f",
|
||||
"8302cd44-569c-4109-83dd-87663d053200",
|
||||
"fb63a999-5aa0-4912-8d06-400f852676cc",
|
||||
"1476c46b-84ce-4151-b777-9dac8b533a8c",
|
||||
"518e4b70-6742-43d2-bd39-90b3af026d89",
|
||||
"9fa543b2-d5c9-47e1-8cba-ed1e1bfbfb86",
|
||||
"5759f1c8-4974-4333-98c4-2ab0aec97918",
|
||||
"77565fbe-27ea-43ee-bcb7-a218b073af9f",
|
||||
"a9f67c53-f22d-4d72-be23-2ef95eb49443",
|
||||
"ff866750-238b-4eb6-a9b7-7cacd8c6b9d0",
|
||||
"8470bf3b-1760-4d4c-8755-387fcc5c92e8",
|
||||
"293b9d4a-be83-4c72-b08e-02d6170ac405",
|
||||
"d00b4e66-dfd1-45e8-af78-b643f59b5327",
|
||||
"de6cbb4d-5225-4975-91f3-23b280492598",
|
||||
"bb5bfca7-38d3-4639-83bd-da639ce2b4b5",
|
||||
"d9a6079c-eb37-47e0-b588-28f87ebc7411",
|
||||
"2d83e2f0-9a87-40a7-89cc-2f0689f52f18",
|
||||
"e985ad1c-5bd1-4a29-b683-f1be109998ba",
|
||||
"081bdfa8-0dfc-43d1-9b85-44cfe72892c1",
|
||||
"79b5af36-6f76-4e35-af5c-ed3ca755b3ac",
|
||||
"9c7d00f4-d307-42f3-b6d4-f95b2a6f7597",
|
||||
"1053c44f-6f09-4a1e-b43f-f858fcb2441c",
|
||||
"5f5005e9-6fe6-435a-992f-46440a803197",
|
||||
"a9fe6c71-e1e1-423f-9811-5318c0b49688",
|
||||
"f06e60ed-6d6d-44ef-b2f9-3ef57dc94cdf"
|
||||
],
|
||||
"seed_jobs_count": 120,
|
||||
"status": "passed",
|
||||
"top_n": 25,
|
||||
"unknown_data_ratio": 0.25,
|
||||
"unknown_fields": {
|
||||
"analysis_features": 0,
|
||||
"description_text": 30,
|
||||
"employment_types": 0,
|
||||
"raw_location": 0
|
||||
}
|
||||
},
|
||||
"status": "passed",
|
||||
"timestamp_utc": "2026-07-29T16:36:43.297692+00:00",
|
||||
"top_n": 25
|
||||
},
|
||||
{
|
||||
"dataset_version": "vr116-2026-07-21",
|
||||
"dedupe": {
|
||||
"case_reports": [
|
||||
{
|
||||
"decision": "exact_external_id",
|
||||
"decision_similarity": 1.0,
|
||||
"expected_match": "seed-infra",
|
||||
"expected_match_present": true,
|
||||
"id": "D-001",
|
||||
"is_expected_match": true,
|
||||
"matched": true,
|
||||
"matched_job_id": "b4c42d9c-22c7-4e89-b88d-b8734e2d7e07",
|
||||
"query": "D-001"
|
||||
},
|
||||
{
|
||||
"decision": "exact_external_id",
|
||||
"decision_similarity": 1.0,
|
||||
"expected_match": "seed-network",
|
||||
"expected_match_present": true,
|
||||
"id": "D-002",
|
||||
"is_expected_match": true,
|
||||
"matched": true,
|
||||
"matched_job_id": "537f63ab-1ea0-45fc-b356-e734a33a99be",
|
||||
"query": "D-002"
|
||||
},
|
||||
{
|
||||
"decision": "new",
|
||||
"decision_similarity": 0.0,
|
||||
"expected_match": null,
|
||||
"expected_match_present": false,
|
||||
"id": "D-003",
|
||||
"is_expected_match": false,
|
||||
"matched": false,
|
||||
"matched_job_id": null,
|
||||
"query": "D-003"
|
||||
},
|
||||
{
|
||||
"decision": "new",
|
||||
"decision_similarity": 0.0,
|
||||
"expected_match": null,
|
||||
"expected_match_present": false,
|
||||
"id": "D-004",
|
||||
"is_expected_match": false,
|
||||
"matched": false,
|
||||
"matched_job_id": null,
|
||||
"query": "D-004"
|
||||
}
|
||||
],
|
||||
"false_merges": 0,
|
||||
"false_negative": 0,
|
||||
"false_positive": 0,
|
||||
"missed_merges": 0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"seed_jobs": 2,
|
||||
"status": "passed",
|
||||
"true_negative": 2,
|
||||
"true_positive": 2
|
||||
},
|
||||
"hardware": {
|
||||
"cpu_count": 20,
|
||||
"debug_mode": true,
|
||||
"platform": "Linux-6.12.54-Unraid-x86_64-with-glibc2.41",
|
||||
"python": "3.13.14",
|
||||
"release": "6.12.54-Unraid"
|
||||
},
|
||||
"jobs_processed": 250,
|
||||
"nfr_007": {
|
||||
"measured": false,
|
||||
"p95_ms": null,
|
||||
"passed": false
|
||||
},
|
||||
"parser": {
|
||||
"cases": [
|
||||
{
|
||||
"coverage_ratio": 1.0,
|
||||
"covered_fields": 6,
|
||||
"expected_fields": 6,
|
||||
"expected_job_count": 1,
|
||||
"expected_parser_key": "jsonld-jobposting",
|
||||
"expected_parser_version": "1.0.0",
|
||||
"extracted": 1,
|
||||
"fixture": "fixtures/pages/sample_jsonld_job.html",
|
||||
"id": "P-001",
|
||||
"parser_key": "jsonld-jobposting",
|
||||
"parser_version": "1.0.0",
|
||||
"passed": true,
|
||||
"warnings": []
|
||||
},
|
||||
{
|
||||
"coverage_ratio": 1.0,
|
||||
"covered_fields": 6,
|
||||
"expected_fields": 6,
|
||||
"expected_job_count": 1,
|
||||
"expected_parser_key": "generic-html",
|
||||
"expected_parser_version": "1.1.0",
|
||||
"extracted": 1,
|
||||
"fixture": "fixtures/pages/sample_generic_job.html",
|
||||
"id": "P-002",
|
||||
"parser_key": "generic-html",
|
||||
"parser_version": "1.1.0",
|
||||
"passed": true,
|
||||
"warnings": []
|
||||
}
|
||||
],
|
||||
"coverage_ratio": 1.0,
|
||||
"covered_fields": 12,
|
||||
"expected_fields": 12,
|
||||
"status": "passed",
|
||||
"unknown_data_ratio": 0.0
|
||||
},
|
||||
"performance": {
|
||||
"avg_scores_detail_ms": 55.155,
|
||||
"avg_scores_import_ms": 40.941,
|
||||
"avg_scores_list_ms": 101.206,
|
||||
"avg_scores_rescore_ms": 16.1,
|
||||
"dashboard": {
|
||||
"detail_p95_ms": 60.679,
|
||||
"detail_query_count": 25,
|
||||
"detail_query_types": {
|
||||
"SELECT": 25
|
||||
},
|
||||
"list_p95_ms": 126.661,
|
||||
"list_query_count": 76,
|
||||
"list_query_types": {
|
||||
"SELECT": 76
|
||||
},
|
||||
"timing_count": 50
|
||||
},
|
||||
"import": {
|
||||
"duration_ms_total": 10235.21,
|
||||
"p50_ms": 33.657,
|
||||
"p95_ms": 68.212,
|
||||
"query_count": 3264,
|
||||
"query_types": {
|
||||
"INSERT": 1003,
|
||||
"RELEASE": 253,
|
||||
"SAVEPOINT": 253,
|
||||
"SELECT": 1256,
|
||||
"UPDATE": 499
|
||||
},
|
||||
"timing_count": 250
|
||||
},
|
||||
"jobs": 1,
|
||||
"performance": {
|
||||
"memory_current_mb": 4.49,
|
||||
"memory_peak_mb": 22.389,
|
||||
"nfr_007": true,
|
||||
"p50_ms": 34.958,
|
||||
"p95_ms": 96.727,
|
||||
"total_duration_ms": 14160.325
|
||||
},
|
||||
"query_count": 3370,
|
||||
"rescore": {
|
||||
"duration_ms_total": 16.1,
|
||||
"p50_ms": 16.1,
|
||||
"p95_ms": 16.1,
|
||||
"query_count": 5,
|
||||
"query_types": {
|
||||
"INSERT": 1,
|
||||
"SELECT": 4
|
||||
},
|
||||
"timing_count": 1
|
||||
},
|
||||
"source_run_id": 1,
|
||||
"top_n": 25
|
||||
},
|
||||
"ranking": {
|
||||
"churn_ratio": 0.0,
|
||||
"post_top_n": [
|
||||
"733b522b-382a-4a20-b046-3174b574a456",
|
||||
"21473419-9a52-465d-90d0-da3dba7d87f4",
|
||||
"e97f2b86-717e-4c9a-9d29-ba40f3198809",
|
||||
"f9ee509b-fc89-4d81-8341-5b4ca48a51a1",
|
||||
"f89072f9-b298-4751-b6d6-a3878e54ef18",
|
||||
"f02c6911-d536-40d1-8d8d-3302392452e7",
|
||||
"fb8f2233-2810-4e0e-9977-bebdcc230aef",
|
||||
"8a71292f-bba7-4166-a7b8-714f9147f343",
|
||||
"9cecf422-fd8e-4b40-9529-8e001af3df8b",
|
||||
"80d26bff-3fae-4222-b6d8-25b47631a38b",
|
||||
"d4475ebf-0969-4023-b6e2-e54df6dac7d8",
|
||||
"0b33971f-c114-42a6-89d4-043055cce4d4",
|
||||
"57c2ebf0-6a3c-4233-b1d4-f15bf3dfe087",
|
||||
"bfe61184-136a-4f3f-8d12-4cee2109aa76",
|
||||
"5bb9b9ea-1227-4bbf-b8e9-15de052ed584",
|
||||
"9f361ba8-1e9e-4134-bda6-a7c78b6c4162",
|
||||
"6b7b0dce-036d-475d-81f3-6ed8fcb98c5f",
|
||||
"6c104672-11a2-4ee1-8200-f558d2d3eb0a",
|
||||
"09b629be-5012-403f-b93f-b212d9f46b4c",
|
||||
"46981d16-38d1-46a6-9755-c6d31b5adaf4",
|
||||
"ae01edcc-fe42-4674-97ed-cdeec737ad9a",
|
||||
"131a755e-4c6e-480c-81c4-6f2fe1afb069",
|
||||
"7eba23c8-02ef-41a0-af54-d24689877f38",
|
||||
"81cf6658-2035-4aba-97d9-4032085fba04",
|
||||
"5a0b33ac-4b1e-454d-94b6-ddf5fc0945aa"
|
||||
],
|
||||
"pre_top_n": [
|
||||
"733b522b-382a-4a20-b046-3174b574a456",
|
||||
"21473419-9a52-465d-90d0-da3dba7d87f4",
|
||||
"e97f2b86-717e-4c9a-9d29-ba40f3198809",
|
||||
"f9ee509b-fc89-4d81-8341-5b4ca48a51a1",
|
||||
"f89072f9-b298-4751-b6d6-a3878e54ef18",
|
||||
"f02c6911-d536-40d1-8d8d-3302392452e7",
|
||||
"fb8f2233-2810-4e0e-9977-bebdcc230aef",
|
||||
"8a71292f-bba7-4166-a7b8-714f9147f343",
|
||||
"9cecf422-fd8e-4b40-9529-8e001af3df8b",
|
||||
"80d26bff-3fae-4222-b6d8-25b47631a38b",
|
||||
"d4475ebf-0969-4023-b6e2-e54df6dac7d8",
|
||||
"0b33971f-c114-42a6-89d4-043055cce4d4",
|
||||
"57c2ebf0-6a3c-4233-b1d4-f15bf3dfe087",
|
||||
"bfe61184-136a-4f3f-8d12-4cee2109aa76",
|
||||
"5bb9b9ea-1227-4bbf-b8e9-15de052ed584",
|
||||
"9f361ba8-1e9e-4134-bda6-a7c78b6c4162",
|
||||
"6b7b0dce-036d-475d-81f3-6ed8fcb98c5f",
|
||||
"6c104672-11a2-4ee1-8200-f558d2d3eb0a",
|
||||
"09b629be-5012-403f-b93f-b212d9f46b4c",
|
||||
"46981d16-38d1-46a6-9755-c6d31b5adaf4",
|
||||
"ae01edcc-fe42-4674-97ed-cdeec737ad9a",
|
||||
"131a755e-4c6e-480c-81c4-6f2fe1afb069",
|
||||
"7eba23c8-02ef-41a0-af54-d24689877f38",
|
||||
"81cf6658-2035-4aba-97d9-4032085fba04",
|
||||
"5a0b33ac-4b1e-454d-94b6-ddf5fc0945aa"
|
||||
],
|
||||
"seed_jobs_count": 120,
|
||||
"status": "passed",
|
||||
"top_n": 25,
|
||||
"unknown_data_ratio": 0.25,
|
||||
"unknown_fields": {
|
||||
"analysis_features": 0,
|
||||
"description_text": 30,
|
||||
"employment_types": 0,
|
||||
"raw_location": 0
|
||||
}
|
||||
},
|
||||
"status": "passed",
|
||||
"timestamp_utc": "2026-07-29T16:37:02.017739+00:00",
|
||||
"top_n": 25
|
||||
},
|
||||
{
|
||||
"dataset_version": "vr116-2026-07-21",
|
||||
"dedupe": {
|
||||
"case_reports": [
|
||||
{
|
||||
"decision": "exact_external_id",
|
||||
"decision_similarity": 1.0,
|
||||
"expected_match": "seed-infra",
|
||||
"expected_match_present": true,
|
||||
"id": "D-001",
|
||||
"is_expected_match": true,
|
||||
"matched": true,
|
||||
"matched_job_id": "bcb6a67c-1945-4006-9f01-b332822877ee",
|
||||
"query": "D-001"
|
||||
},
|
||||
{
|
||||
"decision": "exact_external_id",
|
||||
"decision_similarity": 1.0,
|
||||
"expected_match": "seed-network",
|
||||
"expected_match_present": true,
|
||||
"id": "D-002",
|
||||
"is_expected_match": true,
|
||||
"matched": true,
|
||||
"matched_job_id": "bc389f7a-f049-4daa-b01e-b6bab77c81c9",
|
||||
"query": "D-002"
|
||||
},
|
||||
{
|
||||
"decision": "new",
|
||||
"decision_similarity": 0.0,
|
||||
"expected_match": null,
|
||||
"expected_match_present": false,
|
||||
"id": "D-003",
|
||||
"is_expected_match": false,
|
||||
"matched": false,
|
||||
"matched_job_id": null,
|
||||
"query": "D-003"
|
||||
},
|
||||
{
|
||||
"decision": "new",
|
||||
"decision_similarity": 0.0,
|
||||
"expected_match": null,
|
||||
"expected_match_present": false,
|
||||
"id": "D-004",
|
||||
"is_expected_match": false,
|
||||
"matched": false,
|
||||
"matched_job_id": null,
|
||||
"query": "D-004"
|
||||
}
|
||||
],
|
||||
"false_merges": 0,
|
||||
"false_negative": 0,
|
||||
"false_positive": 0,
|
||||
"missed_merges": 0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"seed_jobs": 2,
|
||||
"status": "passed",
|
||||
"true_negative": 2,
|
||||
"true_positive": 2
|
||||
},
|
||||
"hardware": {
|
||||
"cpu_count": 20,
|
||||
"debug_mode": true,
|
||||
"platform": "Linux-6.12.54-Unraid-x86_64-with-glibc2.41",
|
||||
"python": "3.13.14",
|
||||
"release": "6.12.54-Unraid"
|
||||
},
|
||||
"jobs_processed": 250,
|
||||
"nfr_007": {
|
||||
"measured": false,
|
||||
"p95_ms": null,
|
||||
"passed": false
|
||||
},
|
||||
"parser": {
|
||||
"cases": [
|
||||
{
|
||||
"coverage_ratio": 1.0,
|
||||
"covered_fields": 6,
|
||||
"expected_fields": 6,
|
||||
"expected_job_count": 1,
|
||||
"expected_parser_key": "jsonld-jobposting",
|
||||
"expected_parser_version": "1.0.0",
|
||||
"extracted": 1,
|
||||
"fixture": "fixtures/pages/sample_jsonld_job.html",
|
||||
"id": "P-001",
|
||||
"parser_key": "jsonld-jobposting",
|
||||
"parser_version": "1.0.0",
|
||||
"passed": true,
|
||||
"warnings": []
|
||||
},
|
||||
{
|
||||
"coverage_ratio": 1.0,
|
||||
"covered_fields": 6,
|
||||
"expected_fields": 6,
|
||||
"expected_job_count": 1,
|
||||
"expected_parser_key": "generic-html",
|
||||
"expected_parser_version": "1.1.0",
|
||||
"extracted": 1,
|
||||
"fixture": "fixtures/pages/sample_generic_job.html",
|
||||
"id": "P-002",
|
||||
"parser_key": "generic-html",
|
||||
"parser_version": "1.1.0",
|
||||
"passed": true,
|
||||
"warnings": []
|
||||
}
|
||||
],
|
||||
"coverage_ratio": 1.0,
|
||||
"covered_fields": 12,
|
||||
"expected_fields": 12,
|
||||
"status": "passed",
|
||||
"unknown_data_ratio": 0.0
|
||||
},
|
||||
"performance": {
|
||||
"avg_scores_detail_ms": 50.504,
|
||||
"avg_scores_import_ms": 36.681,
|
||||
"avg_scores_list_ms": 102.301,
|
||||
"avg_scores_rescore_ms": 19.548,
|
||||
"dashboard": {
|
||||
"detail_p95_ms": 54.227,
|
||||
"detail_query_count": 25,
|
||||
"detail_query_types": {
|
||||
"SELECT": 25
|
||||
},
|
||||
"list_p95_ms": 129.536,
|
||||
"list_query_count": 76,
|
||||
"list_query_types": {
|
||||
"SELECT": 76
|
||||
},
|
||||
"timing_count": 50
|
||||
},
|
||||
"import": {
|
||||
"duration_ms_total": 9170.368,
|
||||
"p50_ms": 35.32,
|
||||
"p95_ms": 49.306,
|
||||
"query_count": 3264,
|
||||
"query_types": {
|
||||
"INSERT": 1003,
|
||||
"RELEASE": 253,
|
||||
"SAVEPOINT": 253,
|
||||
"SELECT": 1256,
|
||||
"UPDATE": 499
|
||||
},
|
||||
"timing_count": 250
|
||||
},
|
||||
"jobs": 1,
|
||||
"performance": {
|
||||
"memory_current_mb": 12.332,
|
||||
"memory_peak_mb": 16.658,
|
||||
"nfr_007": true,
|
||||
"p50_ms": 36.098,
|
||||
"p95_ms": 93.787,
|
||||
"total_duration_ms": 13010.048
|
||||
},
|
||||
"query_count": 3370,
|
||||
"rescore": {
|
||||
"duration_ms_total": 19.548,
|
||||
"p50_ms": 19.548,
|
||||
"p95_ms": 19.548,
|
||||
"query_count": 5,
|
||||
"query_types": {
|
||||
"INSERT": 1,
|
||||
"SELECT": 4
|
||||
},
|
||||
"timing_count": 1
|
||||
},
|
||||
"source_run_id": 1,
|
||||
"top_n": 25
|
||||
},
|
||||
"ranking": {
|
||||
"churn_ratio": 0.0,
|
||||
"post_top_n": [
|
||||
"6758530f-71d8-4e3d-9732-7e1001eefd13",
|
||||
"94176301-843f-4074-abdf-fd2c76b76d36",
|
||||
"492a021b-a00d-47ff-b236-86ef6a88a2a6",
|
||||
"c6227dbc-82c4-4c4b-8aec-99714d5adc53",
|
||||
"523a1940-cabc-48b3-abab-bc74f18d4ee4",
|
||||
"ca18c4cc-b796-4ebf-9591-c335bc072f57",
|
||||
"c6a495ac-564a-4aee-a509-081100ddf0a8",
|
||||
"0cf16e0f-5ffc-455b-b488-deb19b0e0fea",
|
||||
"b906c635-a818-4db7-b639-dc6e745ef345",
|
||||
"edce515a-e697-4925-b73f-5605b1236727",
|
||||
"1f800e1f-ff44-4548-91c7-0e9f97af40e2",
|
||||
"0d3caa61-dab1-49d1-8af3-21cb605e9b41",
|
||||
"e6f0d2ea-6dfc-4559-8aa9-80799402e8f1",
|
||||
"a5d682ee-6fac-4ae9-921d-7e883db86d49",
|
||||
"543f63fe-f7fd-4302-9414-95536f5fabcc",
|
||||
"b97df085-3019-45cb-aee6-66edb6602df9",
|
||||
"733c02de-73dc-4967-9e34-c62dcdd98e4f",
|
||||
"f0178164-2535-4e02-91c7-7a6920ab6df3",
|
||||
"4bb41af9-5387-4d0c-a62b-37f7ff713dfe",
|
||||
"ea01b4ee-5a92-4df2-8896-bd391b5c661b",
|
||||
"f08116ab-fa5d-45e7-a579-d9c2cad0d63f",
|
||||
"3c4feccb-a4d1-4cd1-b814-1a2d8bfd9da1",
|
||||
"69088d9a-0de5-4bbe-b8bf-abd733838ed6",
|
||||
"e567d05d-8ed4-4f0f-80fb-1e6f0f77f377",
|
||||
"9f03de2f-4c1e-4cc5-9d65-bbcd3b42a029"
|
||||
],
|
||||
"pre_top_n": [
|
||||
"6758530f-71d8-4e3d-9732-7e1001eefd13",
|
||||
"94176301-843f-4074-abdf-fd2c76b76d36",
|
||||
"492a021b-a00d-47ff-b236-86ef6a88a2a6",
|
||||
"c6227dbc-82c4-4c4b-8aec-99714d5adc53",
|
||||
"523a1940-cabc-48b3-abab-bc74f18d4ee4",
|
||||
"ca18c4cc-b796-4ebf-9591-c335bc072f57",
|
||||
"c6a495ac-564a-4aee-a509-081100ddf0a8",
|
||||
"0cf16e0f-5ffc-455b-b488-deb19b0e0fea",
|
||||
"b906c635-a818-4db7-b639-dc6e745ef345",
|
||||
"edce515a-e697-4925-b73f-5605b1236727",
|
||||
"1f800e1f-ff44-4548-91c7-0e9f97af40e2",
|
||||
"0d3caa61-dab1-49d1-8af3-21cb605e9b41",
|
||||
"e6f0d2ea-6dfc-4559-8aa9-80799402e8f1",
|
||||
"a5d682ee-6fac-4ae9-921d-7e883db86d49",
|
||||
"543f63fe-f7fd-4302-9414-95536f5fabcc",
|
||||
"b97df085-3019-45cb-aee6-66edb6602df9",
|
||||
"733c02de-73dc-4967-9e34-c62dcdd98e4f",
|
||||
"f0178164-2535-4e02-91c7-7a6920ab6df3",
|
||||
"4bb41af9-5387-4d0c-a62b-37f7ff713dfe",
|
||||
"ea01b4ee-5a92-4df2-8896-bd391b5c661b",
|
||||
"f08116ab-fa5d-45e7-a579-d9c2cad0d63f",
|
||||
"3c4feccb-a4d1-4cd1-b814-1a2d8bfd9da1",
|
||||
"69088d9a-0de5-4bbe-b8bf-abd733838ed6",
|
||||
"e567d05d-8ed4-4f0f-80fb-1e6f0f77f377",
|
||||
"9f03de2f-4c1e-4cc5-9d65-bbcd3b42a029"
|
||||
],
|
||||
"seed_jobs_count": 120,
|
||||
"status": "passed",
|
||||
"top_n": 25,
|
||||
"unknown_data_ratio": 0.25,
|
||||
"unknown_fields": {
|
||||
"analysis_features": 0,
|
||||
"description_text": 30,
|
||||
"employment_types": 0,
|
||||
"raw_location": 0
|
||||
}
|
||||
},
|
||||
"status": "passed",
|
||||
"timestamp_utc": "2026-07-29T16:37:19.113439+00:00",
|
||||
"top_n": 25
|
||||
}
|
||||
],
|
||||
"schema_version": "1.0",
|
||||
"status": "passed",
|
||||
"summary": {
|
||||
"duration_ms_mean": 9040.123,
|
||||
"memory_peak_mb_mean": 18.799,
|
||||
"p50_ms_mean": 32.9,
|
||||
"p95_ms_mean": 52.296,
|
||||
"query_count_mean": 3264,
|
||||
"query_count_total_mean": 3370,
|
||||
"throughput_per_second": 28.031
|
||||
},
|
||||
"warmup_status": "passed"
|
||||
}
|
||||
+1
-1
@@ -139,7 +139,7 @@ CSRF_TRUSTED_ORIGINS = merge_unique(
|
||||
)
|
||||
TIME_ZONE = os.getenv("DJANGO_TIME_ZONE", "Europe/Brussels")
|
||||
VACATURERADAR_OWNER_NAME = os.getenv("VACATURERADAR_OWNER_NAME", "Jens").strip()[:40] or "Jens"
|
||||
VACATURERADAR_VERSION = os.getenv("VACATURERADAR_VERSION", "0.3.13").strip() or "0.3.13"
|
||||
VACATURERADAR_VERSION = os.getenv("VACATURERADAR_VERSION", "0.3.14").strip() or "0.3.14"
|
||||
LANGUAGE_CODE = "nl-be"
|
||||
USE_I18N = True
|
||||
USE_TZ = True
|
||||
|
||||
@@ -9,7 +9,7 @@ DJANGO_ALLOWED_HOSTS=vacatureradar.example.be,127.0.0.1,localhost
|
||||
DJANGO_CSRF_TRUSTED_ORIGINS=https://vacatureradar.example.be
|
||||
DJANGO_TIME_ZONE=Europe/Brussels
|
||||
VACATURERADAR_OWNER_NAME=Jens
|
||||
VACATURERADAR_VERSION=0.3.13
|
||||
VACATURERADAR_VERSION=0.3.14
|
||||
|
||||
POSTGRES_DB=vacatureradar
|
||||
POSTGRES_USER=vacatureradar
|
||||
|
||||
@@ -2329,3 +2329,33 @@ tasks:
|
||||
note: Release 0.3.13 professionaliseert responsive layout, mobiele gastflow, begrensde vacature-UX, verplichte
|
||||
zes-viewport-Playwrightgate zonder skips, echte runtimeheartbeat, parserbenchmark, authdekking, SBOM en integrale
|
||||
documentatie; volledige gate 312 tests groen, 84,16% dekking.
|
||||
- id: VR-228
|
||||
title: Optimaliseer de importpipeline voor release 0.3.14
|
||||
status: ready
|
||||
priority: P0
|
||||
requirement_ids:
|
||||
- NFR-007
|
||||
- NFR-008
|
||||
- NFR-009
|
||||
depends_on:
|
||||
- VR-227
|
||||
summary: Profileer de PostgreSQL-import reproduceerbaar, verwijder aantoonbare N+1- en overbodige writes en lever
|
||||
de geverifieerde optimalisatie veilig op zonder wijziging aan deduplicatie, scoring of datakwaliteit.
|
||||
acceptance_criteria:
|
||||
- Een fail-closed runner meet een disposable PostgreSQL-database met vaste fixture, warm-up en drie herhalingen.
|
||||
- De geoptimaliseerde import reduceert queryvolume en doorlooptijd aantoonbaar tegenover 0.3.13 en rapporteert
|
||||
p50, p95, throughput, geheugen en queryfamilies.
|
||||
- Parserdekking blijft 12/12, dedupeprecision en -recall blijven 1,0, rankingchurn blijft nul en historische
|
||||
ScoreRuns plus veldprovenance blijven intact.
|
||||
- Regressietests begrenzen het queryvolume; de volledige kwaliteitsgate slaagt tweemaal en productieacceptatie
|
||||
bewijst health, beperkte importsmoke, publieke kernflows en rollbackgereedheid.
|
||||
verification:
|
||||
- python -m scripts.benchmark_import --jobs 250 --iterations 3
|
||||
- uv run pytest tests/integration/test_pipeline.py tests/unit/test_import_benchmark.py
|
||||
- ./scripts/codex_verify.sh
|
||||
primary_paths:
|
||||
- apps/jobs/services/pipeline.py
|
||||
- scripts/benchmark.py
|
||||
- scripts/benchmark_import.py
|
||||
- tests/integration/test_pipeline.py
|
||||
- docs/audit/IMPORT_PERFORMANCE_FINAL.md
|
||||
|
||||
@@ -0,0 +1,14 @@
|
||||
# Importperformancebaseline 0.3.13
|
||||
|
||||
Gemeten op 2026-07-29 in een disposable PostgreSQL-database op Unraid (Linux 6.12.54, Python 3.13.14, 20 CPU's). De vaste fixture bevat 250 geldige HTML-documenten met dezelfde externe vacature-identiteit: één create en 249 idempotente duplicate-updates. Concurrency is bewust 1; netwerk, Celery, Ollama en mail zijn uit de meting gehouden.
|
||||
|
||||
| Metriek | 0.3.13 |
|
||||
|---|---:|
|
||||
| Importduur | 16.358,158 ms |
|
||||
| p50 per document | 63,034 ms |
|
||||
| p95 per document | 79,871 ms |
|
||||
| Throughput | 15,283 documenten/s |
|
||||
| Importqueries | >9.000 (loggerlimiet bereikt) |
|
||||
| Piekgeheugen | 26,582 MB |
|
||||
|
||||
Correctheidsankers: parserdekking 12/12, dedupeprecision 1,0, deduperecall 1,0 en rankingchurn 0. De overschrijding van 9.000 queries is als ondergrens gerapporteerd, niet als exact getal.
|
||||
@@ -0,0 +1,14 @@
|
||||
# Finale importperformancemeting 0.3.14
|
||||
|
||||
De officiële runner gebruikt een database met verplichte prefix `vacatureradar_bench_`, weigert SQLite, externe databasehosts en publieke productiedomeinen, migreert en flusht de disposable database, voert één warm-up uit en meet daarna drie iteraties van 250 documenten. Het machineleesbare bewijs staat in `artifacts/import-performance-report.json`.
|
||||
|
||||
| Metriek | 0.3.13 | 0.3.14 gemiddeld | Verbetering |
|
||||
|---|---:|---:|---:|
|
||||
| Importduur | 16.358,158 ms | 9.040,123 ms | 44,74% lager |
|
||||
| p50 | 63,034 ms | 32,900 ms | 47,80% lager |
|
||||
| p95 | 79,871 ms | 52,296 ms | 34,53% lager |
|
||||
| Throughput | 15,283/s | 28,031/s | 83,42% hoger |
|
||||
| Importqueries | >9.000 | 3.264 | minstens 63,73% lager |
|
||||
| Piekgeheugen | 26,582 MB | 18,799 MB | 29,28% lager |
|
||||
|
||||
Alle drie iteraties zijn gelijk voor queryvolume en queryfamilies. De kwaliteitsankers bleven parser 12/12, precision 1,0, recall 1,0 en rankingchurn 0. De fixture resulteert per iteratie in één vacature, één versie, behouden veldprovenance, 250 historische ScoreRuns en 249 idempotente duplicates.
|
||||
@@ -0,0 +1,10 @@
|
||||
# Importperformancetraceability
|
||||
|
||||
| Acceptatie | Implementatie | Bewijs |
|
||||
|---|---|---|
|
||||
| Veilige isolatie | `scripts/benchmark_import.py`, `scripts/run_import_benchmark.sh` | negatieve identitytests en disposable PostgreSQL-run |
|
||||
| Meetbare reductie | run-local caches, gebundelde provenance en score-snapshotcopy | baseline/finaal rapport en machine-JSON |
|
||||
| Geen functionele regressie | bestaande dedupe/scoringregels ongewijzigd | pipeline-, dedupe- en benchmarktests |
|
||||
| Historiek en provenance | scorecopy per import, evidence bulk sync | integratietest met 20 herhaalde documenten |
|
||||
| Queryregressie voorkomen | execute-wrapper en budgettest | stabiel 3.264 queries per 250; minder dan 300 per 20 in test |
|
||||
| Releasekwaliteit | VR-228, gates, acceptance en operationsdocs | `docs/ai/PROJECT_STATE.md` en `FINAL_ACCEPTANCE.md` |
|
||||
@@ -0,0 +1,12 @@
|
||||
# Importqueryanalyse 0.3.14
|
||||
|
||||
De hot path liep van `process_raw_document` via parsing en `persist_draft` naar employer-resolutie, deduplicatie, alias/provenance, versiebeheer en scoring. Profiling wees vier dominante oorzaken aan:
|
||||
|
||||
1. Dezelfde werkgever en actieve profielen werden voor elk document opnieuw opgehaald.
|
||||
2. Veldprovenance gebruikte per veld een afzonderlijke `update_or_create`.
|
||||
3. Een inhoudelijk ongewijzigde waarneming schreef opnieuw een versie en berekende alle scores volledig.
|
||||
4. `last_changed` verschoof ten onrechte bij alleen een nieuwe `last_seen`-waarneming.
|
||||
|
||||
De oplossing blijft binnen de modulaire monoliet en transacties: een import-run krijgt een lokale `PersistenceContext`, provenance wordt per alias gelezen en gebundeld geschreven, versies ontstaan alleen bij inhoudelijke wijziging en een ongewijzigde job kopieert het laatste ongewijzigde scoresnapshot. Daardoor blijft iedere `ScoreRun` historisch aanwezig, terwijl deterministische scoring niet opnieuw wordt uitgevoerd. De cache leeft nooit buiten één seriële import-run en verandert workerisolatie of deduplicatieregels niet.
|
||||
|
||||
De database-executieteller rapporteert per 250 items exact: 1.256 SELECT, 1.003 INSERT, 499 UPDATE, 253 SAVEPOINT en 253 RELEASE; totaal 3.264 importexecuties. De drie officiële iteraties geven identieke queryfamilies.
|
||||
@@ -1,5 +1,9 @@
|
||||
# Performanceaudit 2026-07-29
|
||||
|
||||
## Importoptimalisatie 0.3.14
|
||||
|
||||
De eerder geregistreerde importhotspot is opgelost en reproduceerbaar gemeten in een disposable PostgreSQL-database. Tegenover de 0.3.13-baseline daalt de gemiddelde 250-itemimport van 16.358,158 naar 9.040,123 ms; p95 daalt van 79,871 naar 52,296 ms en queryvolume van meer dan 9.000 naar exact 3.264. Zie `IMPORT_PERFORMANCE_BASELINE.md`, `IMPORT_QUERY_ANALYSIS.md` en `IMPORT_PERFORMANCE_FINAL.md`.
|
||||
|
||||
## Baseline
|
||||
|
||||
- Lokale quickbenchmark: 250 invoeritems op Windows 11, Python 3.13, 16 CPU-threads.
|
||||
|
||||
@@ -16,3 +16,4 @@
|
||||
| Back-up/restore | custom dump, media en runbook | geïsoleerde restore, exacte tellingen en aparte applicatiestart | implemented and live verified |
|
||||
| Rollback | immutable vorige en huidige imagetags | oude image healthy, daarna 0.3.13 opnieuw healthy | implemented and live verified |
|
||||
| Externe identity/mail | versleutelde configuratieboundary | `USER_INPUT_REQUIRED.md` | optional external action |
|
||||
| Importperformance 0.3.14 | run-lokale caches, evidence bulk sync en scorecopy | geïsoleerde 3-run PostgreSQL-meting en regressietests | implemented and verified |
|
||||
|
||||
@@ -84,6 +84,7 @@ Legenda: **I** = geïmplementeerd en getest in de huidige MVP; **P** = gedeeltel
|
||||
Iedere backlogtaak die een requirement wijzigt, werkt in dezelfde commit deze matrix, relevante acceptatiecriteria en tests bij. Een status wordt alleen naar **I** gezet wanneer de code én het genoemde bewijs bestaan.
|
||||
|
||||
## Aanvullende traceability-opmerking
|
||||
- VR-228 levert release 0.3.14 met een fail-closed disposable PostgreSQL-benchmark, run-lokale importcaches, gebundelde veldprovenance en ongewijzigde score-snapshotcopy. De officiële drie iteraties bewijzen 3.264 queries per 250 items, parser 12/12, dedupeprecision/recall 1,0 en nul rankingchurn; bewijs staat in `docs/audit/IMPORT_PERFORMANCE_*.md` en het machineleesbare artifact.
|
||||
- VR-227 professionaliseert release 0.3.13 integraal: gecentreerd ultrawideframe, mobiele gastbanner en verticale pipeline, begrensde vacaturepaginering, verplichte zes-viewport-Playwrightgate zonder skips, echte runtimeheartbeat, volledige generieke parserbenchmark en SBOM-generatie. Bewijs staat in `docs/audit/`, de E2E-/auth-/runtime-/adaptertests en `USER_INPUT_REQUIRED.md`.
|
||||
- VR-102 is gerealiseerd met auditbare policyreviews (`SourcePolicyReview`), robotscache met TTL/size/SSRF-controles en fail-closed policy-gating op verlopen of conflicterende reviews.
|
||||
- VR-107 is gerealiseerd met `apps/sources/views.py`, `apps/sources/services/manual_import.py`, `templates/sources/list.html`, `static/js/manual_import.js` en bijhorende unit/integration-tests.
|
||||
|
||||
@@ -0,0 +1,5 @@
|
||||
# VacatureRadar 0.3.14
|
||||
|
||||
Deze release versnelt herhaalde vacature-imports zonder deduplicatie, scoring of gebruikersdata te veranderen. Een geïsoleerde PostgreSQL-benchmark bewijst 44,74% kortere importduur, minstens 63,73% minder databasequeries en 83,42% hogere throughput tegenover 0.3.13. De runner, regressiebudgetten en queryfamilierapportage maken deze winst reproduceerbaar.
|
||||
|
||||
Upgrade vereist geen schemawijziging. Voer zoals altijd `migrate --noinput` uit; rollback kan naar de bewaarde immutable 0.3.13-image zonder dataconversie.
|
||||
+1
-1
@@ -1,6 +1,6 @@
|
||||
[project]
|
||||
name = "vacatureradar"
|
||||
version = "0.3.13"
|
||||
version = "0.3.14"
|
||||
description = "Persoonlijke autonome vacature-assistent voor toegestane publieke bronnen en vacaturemails."
|
||||
readme = "README.md"
|
||||
requires-python = ">=3.12,<3.14"
|
||||
|
||||
+32
-14
@@ -20,7 +20,6 @@ import django
|
||||
from django.conf import settings
|
||||
from django.contrib.auth import get_user_model
|
||||
from django.db import connection
|
||||
from django.test.utils import CaptureQueriesContext
|
||||
from django.utils import timezone
|
||||
|
||||
os.environ.setdefault("DJANGO_SETTINGS_MODULE", "config.settings")
|
||||
@@ -29,7 +28,7 @@ django.setup()
|
||||
from apps.jobs.models import Employer, JobPosting, JobSourceAlias, ScoreRun # noqa: E402
|
||||
from apps.jobs.services.dedupe import find_existing_job # noqa: E402
|
||||
from apps.jobs.services.normalization import CanonicalJobDraft, normalize_token # noqa: E402
|
||||
from apps.jobs.services.pipeline import process_raw_document # noqa: E402
|
||||
from apps.jobs.services.pipeline import PersistenceContext, process_raw_document # noqa: E402
|
||||
from apps.jobs.services.scoring import calculate_score, rescore_jobs_with_profiles # noqa: E402
|
||||
from apps.profiles.models import SearchProfile # noqa: E402
|
||||
from apps.sources.adapters.registry import registry # noqa: E402
|
||||
@@ -575,15 +574,25 @@ def _clean_benchmark(artifacts: BenchmarkArtifact) -> None:
|
||||
get_user_model().objects.filter(pk=artifacts.user_id).delete()
|
||||
|
||||
|
||||
class QueryCounter:
|
||||
"""Count database executions without Django's bounded debug-query log."""
|
||||
|
||||
def __init__(self) -> None:
|
||||
self.count = 0
|
||||
self.types: dict[str, int] = {}
|
||||
|
||||
def __call__(self, execute: Any, sql: str, params: Any, many: bool, context: Any) -> Any:
|
||||
self.count += 1
|
||||
query_type = sql.lstrip().split(None, 1)[0].upper() if sql.strip() else "UNKNOWN"
|
||||
self.types[query_type] = self.types.get(query_type, 0) + 1
|
||||
return execute(sql, params, many, context)
|
||||
|
||||
|
||||
@contextmanager
|
||||
def _query_capture() -> Any:
|
||||
original_force_debug_cursor = connection.force_debug_cursor
|
||||
connection.force_debug_cursor = True
|
||||
try:
|
||||
with CaptureQueriesContext(connection) as captured:
|
||||
yield captured
|
||||
finally:
|
||||
connection.force_debug_cursor = original_force_debug_cursor
|
||||
counter = QueryCounter()
|
||||
with connection.execute_wrapper(counter):
|
||||
yield counter
|
||||
|
||||
|
||||
def run_performance_benchmark(
|
||||
@@ -619,6 +628,7 @@ def run_performance_benchmark(
|
||||
}
|
||||
tracemalloc.start()
|
||||
import_timings: list[float] = []
|
||||
persistence_context = PersistenceContext()
|
||||
rescore_timings: list[float] = []
|
||||
list_timings: list[float] = []
|
||||
detail_timings: list[float] = []
|
||||
@@ -646,9 +656,10 @@ def run_performance_benchmark(
|
||||
retain_until=timezone.now() + timezone.timedelta(days=7),
|
||||
)
|
||||
start = perf_counter()
|
||||
process_raw_document(document)
|
||||
process_raw_document(document, context=persistence_context)
|
||||
import_timings.append((perf_counter() - start) * 1000)
|
||||
import_query_count = len(captured_import.captured_queries)
|
||||
import_query_count = captured_import.count
|
||||
import_query_types = captured_import.types
|
||||
query_count_total += import_query_count
|
||||
|
||||
candidates = list(
|
||||
@@ -666,7 +677,8 @@ def run_performance_benchmark(
|
||||
batch_start = perf_counter()
|
||||
rescore_jobs_with_profiles(batch, profile_id=profile.id)
|
||||
rescore_timings.append((perf_counter() - batch_start) * 1000)
|
||||
rescore_query_count = len(captured_rescore.captured_queries)
|
||||
rescore_query_count = captured_rescore.count
|
||||
rescore_query_types = captured_rescore.types
|
||||
query_count_total += rescore_query_count
|
||||
|
||||
with _query_capture() as captured_list:
|
||||
@@ -695,8 +707,10 @@ def run_performance_benchmark(
|
||||
)
|
||||
detail_timings.append((perf_counter() - start) * 1000)
|
||||
|
||||
list_query_count = len(captured_list.captured_queries)
|
||||
detail_query_count = len(captured_detail.captured_queries)
|
||||
list_query_count = captured_list.count
|
||||
detail_query_count = captured_detail.count
|
||||
list_query_types = captured_list.types
|
||||
detail_query_types = captured_detail.types
|
||||
query_count_total += list_query_count + detail_query_count
|
||||
|
||||
current_bytes, peak_bytes = tracemalloc.get_traced_memory()
|
||||
@@ -740,6 +754,7 @@ def run_performance_benchmark(
|
||||
"p50_ms": timing_stats["import_p50_ms"],
|
||||
"p95_ms": timing_stats["import_p95_ms"],
|
||||
"query_count": import_query_count,
|
||||
"query_types": import_query_types,
|
||||
"timing_count": len(import_timings),
|
||||
},
|
||||
"rescore": {
|
||||
@@ -747,11 +762,14 @@ def run_performance_benchmark(
|
||||
"p50_ms": timing_stats["rescore_p50_ms"],
|
||||
"p95_ms": timing_stats["rescore_p95_ms"],
|
||||
"query_count": rescore_query_count,
|
||||
"query_types": rescore_query_types,
|
||||
"timing_count": len(rescore_timings),
|
||||
},
|
||||
"dashboard": {
|
||||
"list_query_count": list_query_count,
|
||||
"detail_query_count": detail_query_count,
|
||||
"list_query_types": list_query_types,
|
||||
"detail_query_types": detail_query_types,
|
||||
"list_p95_ms": timing_stats["list_p95_ms"],
|
||||
"detail_p95_ms": timing_stats["detail_p95_ms"],
|
||||
"timing_count": len(list_timings) + len(detail_timings),
|
||||
|
||||
@@ -0,0 +1,126 @@
|
||||
#!/usr/bin/env python3
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import os
|
||||
from pathlib import Path
|
||||
from statistics import mean
|
||||
from typing import Any
|
||||
from urllib.parse import urlparse
|
||||
|
||||
os.environ.setdefault("DJANGO_SETTINGS_MODULE", "config.settings")
|
||||
|
||||
import django
|
||||
|
||||
django.setup()
|
||||
|
||||
from django.conf import settings # noqa: E402
|
||||
from django.core.management import call_command # noqa: E402
|
||||
|
||||
from scripts.benchmark import DEFAULT_DATASET, run_benchmark_report # noqa: E402
|
||||
|
||||
SAFE_DATABASE_PREFIX = "vacatureradar_bench_"
|
||||
SAFE_PUBLIC_HOSTS = {"", "benchmark.invalid", "localhost", "127.0.0.1"}
|
||||
|
||||
|
||||
class UnsafeBenchmarkEnvironment(RuntimeError):
|
||||
pass
|
||||
|
||||
|
||||
def validate_benchmark_identity(database: dict[str, Any], public_base_url: str) -> None:
|
||||
engine = str(database.get("ENGINE", ""))
|
||||
name = str(database.get("NAME", ""))
|
||||
host = str(database.get("HOST", ""))
|
||||
public_host = (urlparse(public_base_url).hostname or "").casefold()
|
||||
if engine != "django.db.backends.postgresql":
|
||||
raise UnsafeBenchmarkEnvironment("De importbenchmark vereist PostgreSQL.")
|
||||
if not name.startswith(SAFE_DATABASE_PREFIX):
|
||||
raise UnsafeBenchmarkEnvironment(
|
||||
f"Onveilige databasenaam {name!r}; verwacht prefix {SAFE_DATABASE_PREFIX!r}."
|
||||
)
|
||||
if host not in {"127.0.0.1", "localhost", "postgres"}:
|
||||
raise UnsafeBenchmarkEnvironment(f"Onveilige benchmarkdatabasehost {host!r}.")
|
||||
if public_host not in SAFE_PUBLIC_HOSTS:
|
||||
raise UnsafeBenchmarkEnvironment(
|
||||
f"Publieke host {public_host!r} is niet toegestaan voor een benchmark."
|
||||
)
|
||||
|
||||
|
||||
def _summary(reports: list[dict[str, Any]]) -> dict[str, float]:
|
||||
performance = [report["performance"] for report in reports]
|
||||
imports = [item["import"] for item in performance]
|
||||
return {
|
||||
"query_count_mean": round(mean(item["query_count"] for item in imports), 3),
|
||||
"query_count_total_mean": round(mean(item["query_count"] for item in performance), 3),
|
||||
"duration_ms_mean": round(mean(item["duration_ms_total"] for item in imports), 3),
|
||||
"p50_ms_mean": round(mean(item["p50_ms"] for item in imports), 3),
|
||||
"p95_ms_mean": round(mean(item["p95_ms"] for item in imports), 3),
|
||||
"throughput_per_second": round(
|
||||
mean(
|
||||
report["jobs_processed"] / (item["duration_ms_total"] / 1000)
|
||||
for report, item in zip(reports, imports, strict=True)
|
||||
),
|
||||
3,
|
||||
),
|
||||
"memory_peak_mb_mean": round(
|
||||
mean(item["performance"]["memory_peak_mb"] for item in performance), 3
|
||||
),
|
||||
}
|
||||
|
||||
|
||||
def run_isolated_benchmark(
|
||||
*, dataset: Path, jobs: int, iterations: int, output: Path
|
||||
) -> dict[str, Any]:
|
||||
validate_benchmark_identity(settings.DATABASES["default"], settings.PUBLIC_BASE_URL)
|
||||
if iterations < 1:
|
||||
raise ValueError("iterations moet minstens 1 zijn.")
|
||||
call_command("migrate", interactive=False, verbosity=0)
|
||||
call_command("flush", interactive=False, verbosity=0)
|
||||
warmup = run_benchmark_report(dataset, quick=True, jobs=jobs)
|
||||
measured: list[dict[str, Any]] = []
|
||||
for _ in range(iterations):
|
||||
call_command("flush", interactive=False, verbosity=0)
|
||||
measured.append(run_benchmark_report(dataset, quick=True, jobs=jobs))
|
||||
report = {
|
||||
"schema_version": "1.0",
|
||||
"database_identity": {
|
||||
"engine": settings.DATABASES["default"]["ENGINE"],
|
||||
"name_prefix": SAFE_DATABASE_PREFIX,
|
||||
"host": settings.DATABASES["default"]["HOST"],
|
||||
},
|
||||
"configuration": {
|
||||
"jobs": jobs,
|
||||
"iterations": iterations,
|
||||
"warmup_runs": 1,
|
||||
"concurrency": 1,
|
||||
"dataset": str(dataset),
|
||||
},
|
||||
"warmup_status": warmup["status"],
|
||||
"summary": _summary(measured),
|
||||
"iterations": measured,
|
||||
"status": "passed" if all(item["status"] == "passed" for item in measured) else "failed",
|
||||
}
|
||||
output.parent.mkdir(parents=True, exist_ok=True)
|
||||
output.write_text(json.dumps(report, indent=2, sort_keys=True) + "\n", encoding="utf-8")
|
||||
return report
|
||||
|
||||
|
||||
def main() -> int:
|
||||
parser = argparse.ArgumentParser(description="Fail-closed geïsoleerde importbenchmark")
|
||||
parser.add_argument("--dataset", type=Path, default=DEFAULT_DATASET)
|
||||
parser.add_argument("--jobs", type=int, default=250)
|
||||
parser.add_argument("--iterations", type=int, default=3)
|
||||
parser.add_argument(
|
||||
"--output", type=Path, default=Path("artifacts/import-performance-report.json")
|
||||
)
|
||||
args = parser.parse_args()
|
||||
report = run_isolated_benchmark(
|
||||
dataset=args.dataset, jobs=args.jobs, iterations=args.iterations, output=args.output
|
||||
)
|
||||
print(json.dumps(report, indent=2, sort_keys=True))
|
||||
return 0 if report["status"] == "passed" else 1
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
@@ -0,0 +1,16 @@
|
||||
#!/usr/bin/env bash
|
||||
set -euo pipefail
|
||||
cd "$(dirname "$0")/.."
|
||||
|
||||
: "${DATABASE_URL:?Stel DATABASE_URL in op een afzonderlijke vacatureradar_bench_* PostgreSQL-database.}"
|
||||
export PUBLIC_BASE_URL="${PUBLIC_BASE_URL:-https://benchmark.invalid}"
|
||||
export DJANGO_DEBUG=0
|
||||
export DJANGO_SECRET_KEY="${DJANGO_SECRET_KEY:-benchmark-only-secret-key-not-for-production-000000000000000000}"
|
||||
export CELERY_TASK_ALWAYS_EAGER=1
|
||||
export OLLAMA_ENABLED=0
|
||||
export EMAIL_BACKEND=django.core.mail.backends.locmem.EmailBackend
|
||||
|
||||
if command -v uv >/dev/null 2>&1; then
|
||||
exec uv run python -m scripts.benchmark_import "$@"
|
||||
fi
|
||||
exec .venv/bin/python -m scripts.benchmark_import "$@"
|
||||
@@ -3,10 +3,19 @@ from datetime import timedelta
|
||||
from pathlib import Path
|
||||
|
||||
import pytest
|
||||
from django.db import connection
|
||||
from django.test.utils import CaptureQueriesContext
|
||||
from django.utils import timezone
|
||||
|
||||
from apps.jobs.models import Employer, JobPosting, JobSourceAlias, ScoreRun
|
||||
from apps.jobs.services.pipeline import process_raw_document
|
||||
from apps.jobs.models import (
|
||||
Employer,
|
||||
FieldProvenance,
|
||||
JobPosting,
|
||||
JobSourceAlias,
|
||||
JobVersion,
|
||||
ScoreRun,
|
||||
)
|
||||
from apps.jobs.services.pipeline import PersistenceContext, process_raw_document
|
||||
from apps.sources.models import RawDocument, Source
|
||||
|
||||
|
||||
@@ -34,6 +43,39 @@ def test_pipeline_is_idempotent_and_scores(source, profile):
|
||||
assert ScoreRun.objects.filter(profile=profile).count() == 2
|
||||
|
||||
|
||||
@pytest.mark.integration
|
||||
@pytest.mark.django_db
|
||||
def test_unchanged_import_batch_has_bounded_lookup_queries(source, profile):
|
||||
content = Path("fixtures/pages/sample_jsonld_job.html").read_text(encoding="utf-8")
|
||||
documents = [
|
||||
RawDocument.objects.create(
|
||||
source=source,
|
||||
url=f"https://jobs.example.org/import/{index}",
|
||||
final_url=f"https://jobs.example.org/import/{index}",
|
||||
kind=RawDocument.Kind.HTML,
|
||||
content_type="text/html",
|
||||
content_hash=hashlib.sha256(f"{index}{content}".encode()).hexdigest(),
|
||||
body_text=content,
|
||||
byte_length=len(content.encode()),
|
||||
retain_until=timezone.now() + timedelta(days=7),
|
||||
)
|
||||
for index in range(20)
|
||||
]
|
||||
context = PersistenceContext()
|
||||
|
||||
with CaptureQueriesContext(connection) as queries:
|
||||
results = [process_raw_document(document, context=context) for document in documents]
|
||||
|
||||
assert sum(int(result["created"]) for result in results) == 1
|
||||
assert sum(int(result["duplicates"]) for result in results) == 19
|
||||
assert len(queries) < 300
|
||||
assert Employer.objects.count() == 1
|
||||
assert JobPosting.objects.count() == 1
|
||||
assert JobVersion.objects.count() == 1
|
||||
assert FieldProvenance.objects.count() > 0
|
||||
assert ScoreRun.objects.filter(profile=profile).count() == 20
|
||||
|
||||
|
||||
@pytest.mark.django_db
|
||||
def test_pipeline_uses_reviewed_employer_identity_for_public_ats_feed(profile):
|
||||
source = Source.objects.create(
|
||||
|
||||
@@ -0,0 +1,41 @@
|
||||
from pathlib import Path
|
||||
|
||||
import pytest
|
||||
|
||||
from scripts.benchmark_import import UnsafeBenchmarkEnvironment, validate_benchmark_identity
|
||||
|
||||
|
||||
def _database(name: str, host: str = "127.0.0.1") -> dict[str, str]:
|
||||
return {
|
||||
"ENGINE": "django.db.backends.postgresql",
|
||||
"NAME": name,
|
||||
"HOST": host,
|
||||
}
|
||||
|
||||
|
||||
def test_benchmark_identity_accepts_only_isolated_postgres() -> None:
|
||||
validate_benchmark_identity(
|
||||
_database("vacatureradar_bench_release_0314"), "https://benchmark.invalid"
|
||||
)
|
||||
|
||||
|
||||
@pytest.mark.parametrize(
|
||||
("database", "public_url"),
|
||||
[
|
||||
(_database("vacatureradar"), "https://benchmark.invalid"),
|
||||
(
|
||||
_database("vacatureradar_bench_release", "database.internal"),
|
||||
"https://benchmark.invalid",
|
||||
),
|
||||
(_database("vacatureradar_bench_release"), "https://vacatureradar.itworx.tech"),
|
||||
(
|
||||
{"ENGINE": "django.db.backends.sqlite3", "NAME": Path("bench.sqlite3"), "HOST": ""},
|
||||
"https://benchmark.invalid",
|
||||
),
|
||||
],
|
||||
)
|
||||
def test_benchmark_identity_rejects_unsafe_targets(
|
||||
database: dict[str, object], public_url: str
|
||||
) -> None:
|
||||
with pytest.raises(UnsafeBenchmarkEnvironment):
|
||||
validate_benchmark_identity(database, public_url)
|
||||
Reference in New Issue
Block a user