Python: VolatileMemoryStore.get_nearest_matches raises ValueError on an empty collection #14534

What happens

VolatileMemoryStore.get_nearest_matches raises ValueError: cannot reshape array of size 0 into shape (0,newaxis) when the collection exists but has no records. It should return [], as it already does when the collection does not exist.

Where

python/semantic_kernel/memory/volatile_memory_store.py, get_nearest_matches (checked against current main, 9974625):

memory_records = list(self._store[collection_name].values())

# Convert the collection of embeddings into a numpy array (stacked)
embeddings = array([x._embedding for x in memory_records], dtype=float)
embeddings = embeddings.reshape(embeddings.shape[0], -1)

With no records, array([]) has shape (0,) and the reshape to (0, -1) is ambiguous, so numpy raises. Even without the reshape, compute_similarity_scores would raise its own ValueError for "zero vectors" because valid_indices.any() is False for an empty array.

Steps to reproduce

import asyncio, numpy as np
from semantic_kernel.memory.volatile_memory_store import VolatileMemoryStore

async def main():
    s = VolatileMemoryStore()
    await s.create_collection("c")
    print(await s.get_nearest_matches("c", np.array([1.0, 2.0]), limit=3))

asyncio.run(main())

Actual:

  File ".../semantic_kernel/memory/volatile_memory_store.py", line 256, in get_nearest_matches
    embeddings = embeddings.reshape(embeddings.shape[0], -1)
ValueError: cannot reshape array of size 0 into shape (0,newaxis)

Expected: [].

Impact

SemanticTextMemory.search and TextMemoryPlugin.recall call this method. A freshly created collection, or one whose records were all removed, makes a normal search crash instead of returning no results.

I have a fix ready (early return when the collection has no records) with tests.