119 lines
5 KiB
Python
119 lines
5 KiB
Python
import tempfile
|
|
import unittest
|
|
from pathlib import Path
|
|
|
|
from app.services.dataset_export import (
|
|
APPROVED_EXPORT_STATUS,
|
|
DRY_RUN_EXPORT_STATUS,
|
|
ExportKeyMaps,
|
|
build_dataset_record,
|
|
build_manifest,
|
|
cohen_kappa,
|
|
intraclass_correlation,
|
|
scan_for_pii,
|
|
sha256_file,
|
|
write_jsonl,
|
|
)
|
|
|
|
|
|
class DatasetExportTests(unittest.TestCase):
|
|
def test_build_dataset_record_uses_masked_text_and_pseudonymous_keys(self) -> None:
|
|
keys = ExportKeyMaps()
|
|
record = build_dataset_record(
|
|
{
|
|
"session_id": "11111111-1111-1111-1111-111111111111",
|
|
"learner_id": "22222222-2222-2222-2222-222222222222",
|
|
"turn_id": "33333333-3333-3333-3333-333333333333",
|
|
"persona_code": "P1",
|
|
"stage": "rapport",
|
|
"speaker": "counselor",
|
|
"text": "raw text should never be exported",
|
|
"text_masked": "안녕하세요, [NAME]님.",
|
|
"feedback_scores": [{"dimension": "공감", "score": 4}],
|
|
},
|
|
item_index=1,
|
|
export_manifest_id="phase3-rl-seed-test",
|
|
keys=keys,
|
|
)
|
|
|
|
self.assertEqual(record["participant_key"], "PX-0001")
|
|
self.assertEqual(record["session_key"], "SX-0001")
|
|
self.assertEqual(record["turn_key"], "TX-000001")
|
|
self.assertEqual(record["text_masked"], "안녕하세요, [NAME]님.")
|
|
blob = str(record)
|
|
self.assertNotIn("11111111-1111-1111-1111-111111111111", blob)
|
|
self.assertNotIn("22222222-2222-2222-2222-222222222222", blob)
|
|
self.assertNotIn("raw text should never be exported", blob)
|
|
|
|
def test_pii_scan_detects_identifiers_without_raw_samples(self) -> None:
|
|
findings = scan_for_pii(
|
|
{
|
|
"text_masked": "메일 learner@hs.ac.kr, 전화 010-1234-5678, 주민 990101-1234567",
|
|
"api_key": "sk-should-not-appear",
|
|
}
|
|
)
|
|
|
|
kinds = {finding["kind"] for finding in findings}
|
|
self.assertIn("email", kinds)
|
|
self.assertIn("phone", kinds)
|
|
self.assertIn("national_id", kinds)
|
|
self.assertIn("blocked_field", kinds)
|
|
self.assertFalse(any("learner@hs.ac.kr" in finding["sample"] for finding in findings))
|
|
self.assertFalse(any("010-1234-5678" in finding["sample"] for finding in findings))
|
|
|
|
def test_agreement_metrics(self) -> None:
|
|
annotations = [
|
|
{"item_id": "1", "labels": {"appropriateness": "good", "rapport_signal": 4.0}},
|
|
{"item_id": "1", "labels": {"appropriateness": "good", "rapport_signal": 4.1}},
|
|
{"item_id": "2", "labels": {"appropriateness": "bad", "rapport_signal": 2.0}},
|
|
{"item_id": "2", "labels": {"appropriateness": "bad", "rapport_signal": 2.1}},
|
|
{"item_id": "3", "labels": {"appropriateness": "good", "rapport_signal": 5.0}},
|
|
{"item_id": "3", "labels": {"appropriateness": "good", "rapport_signal": 5.0}},
|
|
]
|
|
|
|
self.assertEqual(cohen_kappa(annotations, "appropriateness"), 1.0)
|
|
self.assertGreater(intraclass_correlation(annotations, "rapport_signal") or 0, 0.9)
|
|
|
|
def test_manifest_gate_blocks_unapproved_approved_status(self) -> None:
|
|
with self.assertRaisesRegex(ValueError, "PII scan must pass"):
|
|
build_manifest(
|
|
export_id="phase3-rl-seed-test",
|
|
dataset_name="vignette_phase3_recursive_learning_seed",
|
|
export_status=APPROVED_EXPORT_STATUS,
|
|
purpose="test",
|
|
records=[],
|
|
jsonl_path="03-export/anonymized_dataset.jsonl",
|
|
jsonl_sha256="",
|
|
pii_findings=[{"kind": "email", "path": "$.text_masked", "sample": "<email:hs.ac.kr>"}],
|
|
participants_included=0,
|
|
agreement={"kappa": 0.59, "icc": 0.74, "gold_status": "not_gold"},
|
|
)
|
|
|
|
def test_jsonl_hash_manifest_dry_run(self) -> None:
|
|
record = {
|
|
"schema": "phase3_dataset_item_v1",
|
|
"item_id": "DI-000001",
|
|
"source_refs": {"session_started_at": "2026-06-27T00:00:00Z"},
|
|
}
|
|
with tempfile.TemporaryDirectory() as tmp:
|
|
path = Path(tmp) / "03-export" / "anonymized_dataset.jsonl"
|
|
write_jsonl([record], path)
|
|
digest = sha256_file(path)
|
|
manifest = build_manifest(
|
|
export_id="phase3-rl-seed-test",
|
|
dataset_name="vignette_phase3_recursive_learning_seed",
|
|
export_status=DRY_RUN_EXPORT_STATUS,
|
|
purpose="test",
|
|
records=[record],
|
|
jsonl_path="03-export/anonymized_dataset.jsonl",
|
|
jsonl_sha256=digest,
|
|
pii_findings=[],
|
|
participants_included=1,
|
|
)
|
|
self.assertEqual(manifest["files"][0]["rows"], 1)
|
|
self.assertEqual(manifest["files"][0]["sha256"], digest)
|
|
self.assertEqual(manifest["pii_scan"]["status"], "pass")
|
|
|
|
|
|
if __name__ == "__main__":
|
|
unittest.main()
|