mirror of
https://github.com/supermemoryai/supermemory.git
synced 2026-10-02 02:11:20 +00:00
3d1980560e
`_field` returns the first value that is not None. A v4 hybrid-search hit
shaped `{"memory": "", "chunk": "..."}` therefore resolves to the empty
string and never falls through to `chunk` — the exact fallback the call
site's comment says it is there for.
The consequences are in both directions. In `deduplicate_memories` the hit
produces an empty comparison key and is dropped, so the memory never reaches
the prompt. When such an item does survive, `format_memories_to_text` renders
it through the same `_field` call and emits a bare `- [16 hrs ago] `.
The TypeScript `getMemoryText` takes the first non-empty field instead, and
`agent-framework-python` and `openai-sdk-python` both already match it; the
two voice SDKs were the only implementations that disagreed. Add a
`_memory_text` helper mirroring the TypeScript semantics and use it on both
the dedup and the render path.
Also add tests/conftest.py to pipecat so the import stubs are installed
regardless of collection order — test_utils.py cannot import the package
on its own otherwise.
92 lines
3 KiB
Python
92 lines
3 KiB
Python
from __future__ import annotations
|
|
|
|
import unittest
|
|
from types import SimpleNamespace
|
|
|
|
from supermemory_pipecat.utils import (
|
|
deduplicate_memories,
|
|
format_memories_to_text,
|
|
)
|
|
|
|
|
|
class TestSearchResultExtraction(unittest.TestCase):
|
|
"""v4 hybrid search returns chunk hits whose `memory` field is empty.
|
|
|
|
`_field` stops at the first value that is not None, so those hits used to
|
|
resolve to "" and were dropped by dedup -- or, when they survived, rendered
|
|
as an empty bullet. The TypeScript `getMemoryText` takes the first
|
|
non-empty field instead, and the other Supermemory Python SDKs match it.
|
|
"""
|
|
|
|
def test_falls_through_to_chunk_when_memory_is_empty(self) -> None:
|
|
result = deduplicate_memories(
|
|
static=[],
|
|
dynamic=[],
|
|
search_results=[{"memory": "", "chunk": "The user's dog is called Rex"}],
|
|
)
|
|
|
|
self.assertEqual(len(result["search_results"]), 1)
|
|
self.assertIn("The user's dog is called Rex", format_memories_to_text(result))
|
|
|
|
def test_falls_through_to_chunk_when_memory_is_whitespace(self) -> None:
|
|
result = deduplicate_memories(
|
|
static=[],
|
|
dynamic=[],
|
|
search_results=[{"memory": " ", "chunk": "Chunk body"}],
|
|
)
|
|
|
|
self.assertEqual(len(result["search_results"]), 1)
|
|
self.assertIn("Chunk body", format_memories_to_text(result))
|
|
|
|
def test_falls_through_on_models_too(self) -> None:
|
|
result = deduplicate_memories(
|
|
static=[],
|
|
dynamic=[],
|
|
search_results=[SimpleNamespace(memory="", chunk="Chunk body")],
|
|
)
|
|
|
|
self.assertEqual(len(result["search_results"]), 1)
|
|
self.assertIn("Chunk body", format_memories_to_text(result))
|
|
|
|
def test_never_renders_an_empty_bullet(self) -> None:
|
|
result = deduplicate_memories(
|
|
static=[],
|
|
dynamic=[],
|
|
search_results=[
|
|
{
|
|
"memory": "",
|
|
"chunk": "Chunk body",
|
|
"updatedAt": "2020-01-01T00:00:00Z",
|
|
}
|
|
],
|
|
)
|
|
|
|
rendered = format_memories_to_text(result)
|
|
self.assertIn("Chunk body", rendered)
|
|
for line in rendered.splitlines():
|
|
if line.startswith("- "):
|
|
self.assertNotRegex(line, r"^- (\[[^\]]*\] )?$")
|
|
|
|
def test_memory_still_wins_over_chunk(self) -> None:
|
|
result = deduplicate_memories(
|
|
static=[],
|
|
dynamic=[],
|
|
search_results=[{"memory": "Memory body", "chunk": "Chunk body"}],
|
|
)
|
|
|
|
rendered = format_memories_to_text(result)
|
|
self.assertIn("Memory body", rendered)
|
|
self.assertNotIn("Chunk body", rendered)
|
|
|
|
def test_entry_with_no_usable_text_is_still_dropped(self) -> None:
|
|
result = deduplicate_memories(
|
|
static=[],
|
|
dynamic=[],
|
|
search_results=[{"memory": "", "chunk": ""}, {}, None],
|
|
)
|
|
|
|
self.assertEqual(result["search_results"], [])
|
|
|
|
|
|
if __name__ == "__main__":
|
|
unittest.main()
|