107 lines
3.1 KiB
Python
107 lines
3.1 KiB
Python
from datetime import UTC, datetime
|
|
|
|
import pytest
|
|
from pydantic import ValidationError
|
|
|
|
from tht.corpus.models import CanonicalChunk, CanonicalDocument, CorpusManifest
|
|
|
|
|
|
def document(source_uri: str = "https://host/a.md") -> CanonicalDocument:
|
|
return CanonicalDocument(
|
|
document_id="doc:abc",
|
|
source_id="source:a",
|
|
source_uri=source_uri,
|
|
source_fingerprint="etag:abc",
|
|
content_hash="sha256:def",
|
|
title="A",
|
|
content="# A",
|
|
media_type="text/markdown",
|
|
pipeline_version="normalize-v1",
|
|
)
|
|
|
|
|
|
def chunk() -> CanonicalChunk:
|
|
return CanonicalChunk(
|
|
chunk_id="chunk:abc:0",
|
|
document_id="doc:abc",
|
|
ordinal=0,
|
|
content="# A",
|
|
content_hash="sha256:def",
|
|
source_uri="https://host/a.md",
|
|
pipeline_version="chunk-v1",
|
|
)
|
|
|
|
|
|
def test_manifest_contains_provenance_without_credentials():
|
|
manifest = CorpusManifest(
|
|
manifest_id="manifest:abc",
|
|
created_at=datetime(2026, 7, 12, tzinfo=UTC),
|
|
pipeline_version="evidence-v1",
|
|
embedding_model="nomic-embed-text",
|
|
embedding_dimensions=768,
|
|
documents=[document()],
|
|
chunks=[chunk()],
|
|
)
|
|
|
|
payload = manifest.model_dump_json()
|
|
|
|
assert "https://host/a.md" in payload
|
|
assert "etag:abc" in payload
|
|
assert "evidence-v1" in payload
|
|
assert "nomic-embed-text" in payload
|
|
assert "api_key" not in payload
|
|
|
|
|
|
def test_manifest_can_be_assembled_before_publish_identifiers_are_assigned():
|
|
manifest = CorpusManifest(documents=[document()])
|
|
|
|
assert manifest.documents[0].source_uri == "https://host/a.md"
|
|
assert manifest.manifest_id is None
|
|
|
|
|
|
@pytest.mark.parametrize("model", [document(), chunk()])
|
|
def test_canonical_records_are_frozen(model):
|
|
with pytest.raises(ValidationError):
|
|
model.pipeline_version = "changed" # type: ignore[misc]
|
|
|
|
|
|
def test_manifest_collections_have_independent_defaults():
|
|
first = CorpusManifest(
|
|
manifest_id="one", created_at=datetime.now(UTC), pipeline_version="v1"
|
|
)
|
|
second = CorpusManifest(
|
|
manifest_id="two", created_at=datetime.now(UTC), pipeline_version="v1"
|
|
)
|
|
|
|
first.documents.append(document())
|
|
|
|
assert second.documents == []
|
|
|
|
|
|
def test_manifest_validates_embedding_compatibility_fields():
|
|
with pytest.raises(ValidationError):
|
|
CorpusManifest(
|
|
manifest_id="bad",
|
|
created_at=datetime.now(UTC),
|
|
pipeline_version="v1",
|
|
embedding_dimensions=0,
|
|
)
|
|
|
|
|
|
def test_canonical_metadata_rejects_secrets_and_non_json_values():
|
|
with pytest.raises(ValidationError, match="credential-like"):
|
|
CanonicalDocument.model_validate(
|
|
{**document().model_dump(), "metadata": {"password": "secret"}}
|
|
)
|
|
with pytest.raises(ValidationError):
|
|
CanonicalChunk(
|
|
chunk_id="c",
|
|
document_id="d",
|
|
ordinal=0,
|
|
content="x",
|
|
content_hash="sha256:x",
|
|
source_uri="file:///x",
|
|
pipeline_version="v1",
|
|
metadata={"bad": object()},
|
|
)
|