123 lines
4.3 KiB
Python
123 lines
4.3 KiB
Python
"""L1: tht memory save-one -- targeted upsert via the writer key (spec D11).
|
|
|
|
The D11 deviation: instead of a full vectorstore resync (tht memory index / sync),
|
|
the workflow can push a SINGLE promoted decision to the configured semantic store as
|
|
a one-row upsert. This test pins the pure core of that behavior:
|
|
- exactly one VectorRecord is built for the chosen decision_seq
|
|
- the writer.upsert_records is called once with a single row
|
|
- writer.sync is NEVER called (that is the full-resync path)
|
|
"""
|
|
from datetime import UTC, datetime
|
|
from unittest.mock import MagicMock
|
|
|
|
from tht.adapters.vector.qdrant import point_id
|
|
from tht.memory import MemoryRecord, memory_vector_record_for_decision, save_one_memory
|
|
from tht.vectorstore.records import qdrant_payload
|
|
|
|
|
|
def _record(seq: int = 7, **kw) -> MemoryRecord:
|
|
base = {
|
|
"id": "mem-0007",
|
|
"ts": datetime(2025, 1, 1, tzinfo=UTC),
|
|
"session_id": "s1",
|
|
"decision_seq": seq,
|
|
"type": "concept_clarified",
|
|
"subject": "paziente attivo",
|
|
"detail": "flag_attivo = TRUE",
|
|
"rationale": "r",
|
|
"question_context": "dammi i pazienti",
|
|
"tables": [],
|
|
"concepts": ["paziente attivo"],
|
|
}
|
|
base.update(kw)
|
|
return MemoryRecord(**base)
|
|
|
|
|
|
# --- memory_vector_record_for_decision (single-record filter) ------------------
|
|
|
|
def test_single_record_built_for_decision_seq():
|
|
records = [_record(seq=7), _record(seq=8, id="mem-0008")]
|
|
vr = memory_vector_record_for_decision(records, decision_seq=7)
|
|
assert vr is not None
|
|
assert vr.id == "memory:mem-0007" # memory_vector_records prefix
|
|
assert vr.kind == "memory"
|
|
assert "pazienti" in vr.content
|
|
|
|
|
|
def test_returns_none_for_unknown_decision_seq():
|
|
records = [_record(seq=7)]
|
|
assert memory_vector_record_for_decision(records, decision_seq=999) is None
|
|
|
|
|
|
# --- save_one_memory (the D11 orchestrator: single upsert, never sync) ---------
|
|
|
|
def test_save_one_calls_upsert_with_single_row_never_sync():
|
|
records = [_record(seq=7)]
|
|
writer = MagicMock()
|
|
writer.upsert.return_value = 1
|
|
embedder = MagicMock()
|
|
embedder.embed_documents.return_value = [[0.1] * 8]
|
|
|
|
upserted = save_one_memory(records, decision_seq=7, store=writer, embedder=embedder)
|
|
|
|
assert upserted == 1
|
|
writer.sync.assert_not_called() # the whole point of D11: no full resync
|
|
writer.upsert.assert_called_once()
|
|
args = writer.upsert.call_args
|
|
# table is memory, exactly one row
|
|
assert args[0][0] == "memory"
|
|
rows = args[0][1]
|
|
assert len(rows) == 1
|
|
assert rows[0].record.id == "memory:mem-0007"
|
|
assert rows[0].embedding
|
|
|
|
|
|
def test_save_one_no_record_for_seq_is_noop():
|
|
records = [_record(seq=7)]
|
|
writer = MagicMock()
|
|
embedder = MagicMock()
|
|
upserted = save_one_memory(records, decision_seq=42, store=writer, embedder=embedder)
|
|
assert upserted == 0
|
|
writer.upsert.assert_not_called()
|
|
writer.sync.assert_not_called()
|
|
embedder.embed_documents.assert_not_called()
|
|
|
|
|
|
def test_save_one_uses_writer_key_for_upsert():
|
|
"""The upsert must flow through the writer client (writer key), not a reader.
|
|
Verified indirectly: save_one_memory takes the writer as its client argument."""
|
|
records = [_record(seq=7)]
|
|
writer = MagicMock()
|
|
writer.upsert.return_value = 1
|
|
embedder = MagicMock()
|
|
embedder.embed_documents.return_value = [[0.0] * 4]
|
|
save_one_memory(records, decision_seq=7, store=writer, embedder=embedder)
|
|
# one upsert call, single row, table=memory
|
|
assert writer.upsert.call_count == 1
|
|
|
|
|
|
def test_save_one_preserves_semantic_point_identity_fields():
|
|
records = [_record(seq=7)]
|
|
writer = MagicMock()
|
|
writer.existing_hashes.return_value = {}
|
|
writer.upsert.return_value = 1
|
|
embedder = MagicMock()
|
|
embedder.embed_documents.return_value = [[0.0] * 4]
|
|
|
|
save_one_memory(records, decision_seq=7, store=writer, embedder=embedder)
|
|
|
|
row = writer.upsert.call_args.args[1][0]
|
|
payload = qdrant_payload(
|
|
row.record,
|
|
content_hash=row.content_hash,
|
|
workspace_id="psd-clinical",
|
|
workspace_revision="a" * 40,
|
|
)
|
|
|
|
assert point_id("psd-clinical", "memory", row.record.id) == point_id(
|
|
"psd-clinical", "memory", "memory:mem-0007"
|
|
)
|
|
assert payload["kind"] == "memory"
|
|
assert payload["workspace_id"] == "psd-clinical"
|
|
assert payload["workspace_revision"] == "a" * 40
|