What causes "Persist directory is not set" in microsoft/semantic-kernel?

Pass a valid persist_directory to the constructor.

microsoft/semantic-kernel · error · ServiceInitializationError

Persist directory is not set

Error message

Persist directory is not set

What it means

Raised by USearchMemoryStore._get_all_storage_files (ServiceInitializationError) when _persist_directory is None. This method enumerates on-disk collection files; without a directory there is nothing to scan. It is called during _read_collections_from_dir / re-resolution, so it effectively means 'persistence was never configured but on-disk collection discovery was attempted'.

Source

Thrown at python/semantic_kernel/connectors/memory_stores/usearch/usearch_memory_store.py:543

                    filtered_vectors,
                )
            )
        ]

    def _get_all_storage_files(self) -> dict[str, list[Path]]:
        """Return storage files for each collection in `self._persist_directory`.

        Collection name is derived from file name and converted to lowercase. Files with extensions that
        do not match storage extensions are discarded.

        Raises:
            ValueError: If persist directory is not set.

        Returns:
            Dict[str, List[Path]]: Dictionary of collection names mapped to their respective files.
        """
        if self._persist_directory is None:
            raise ServiceInitializationError("Persist directory is not set")

        storage_exts = _collection_file_extensions.values()
        collection_storage_files: dict[str, list[Path]] = {}
        for path in self._persist_directory.iterdir():
            if path.is_file() and (path.suffix in storage_exts):
                collection_name = path.stem.lower()
                if collection_name in collection_storage_files:
                    collection_storage_files[collection_name].append(path)
                else:
                    collection_storage_files[collection_name] = [path]
        return collection_storage_files

    def _dump_collections(self) -> None:
        collection_storage_files = self._get_all_storage_files()
        for file_path in itertools.chain.from_iterable(collection_storage_files.values()):
            file_path.unlink()

        for collection_name, ucollection in self._collections.items():

View on GitHub (pinned to c028a0c7dc)

Solutions

Pass a valid persist_directory to the constructor.
Ensure the directory exists and is readable.
Avoid code paths that require on-disk discovery for in-memory stores.

Example fix

// before
store = USearchMemoryStore()
// after
store = USearchMemoryStore(persist_directory='/data/usearch')

Defensive patterns

Strategy: validation

Validate before calling

from pathlib import Path
d = Path('/data/usearch')
assert d.exists(), 'persist directory required for storage discovery'
store = USearchMemoryStore(persist_directory=d)

Try / catch

from semantic_kernel.exceptions import ServiceInitializationError
try:
    store = USearchMemoryStore(persist_directory=d)
except ServiceInitializationError:
    store = USearchMemoryStore()

Prevention

Always configure persist_directory when on-disk discovery is needed.
Verify the directory is readable before constructing the store.
Keep persistence configuration consistent across the process lifecycle.

When it happens

Trigger: Constructing `USearchMemoryStore()` without persist_directory and then invoking any internal path that re-scans storage files (e.g. reloading collections). In normal construction, the directory scan is skipped when _persist_directory is falsy, so this is reached via later/direct calls.

Common situations: In-memory store used in a context that later tries to discover collections from disk; misconfigured deployment missing the directory env var; subclass or framework hook calling storage discovery.

Related errors

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/05bdc94f8669760c. Report an issue: GitHub.