Storage API Reference

Complete API reference for all storage protocols and implementations in anchor.

For a usage guide, see Storage Guide.

Protocols

VectorStore

Protocol for vector similarity search backends (FAISS, Chroma, Qdrant, etc.).

class VectorStore(Protocol):
    def add_embedding(
        self, item_id: str, embedding: list[float], metadata: dict[str, Any] | None = None
    ) -> None: ...

    def search(
        self, query_embedding: list[float], top_k: int = 10
    ) -> list[tuple[str, float]]: ...

    def delete(self, item_id: str) -> bool: ...

Method	Description
`add_embedding(item_id, embedding, metadata)`	Store an embedding vector. Overwrites if `item_id` exists.
`search(query_embedding, top_k)`	Return `(item_id, score)` tuples sorted by descending similarity.
`delete(item_id)`	Remove an embedding. Returns `True` if found.

ContextStore

Protocol for storing and retrieving ContextItem objects.

class ContextStore(Protocol):
    def add(self, item: ContextItem) -> None: ...
    def get(self, item_id: str) -> ContextItem | None: ...
    def get_all(self) -> list[ContextItem]: ...
    def delete(self, item_id: str) -> bool: ...
    def clear(self) -> None: ...

Method	Description
`add(item)`	Persist a context item. Overwrites if same `id` exists.
`get(item_id)`	Retrieve by ID. Returns `None` if not found.
`get_all()`	Return all items. Order is implementation-defined.
`delete(item_id)`	Remove by ID. Returns `True` if found.
`clear()`	Remove all items.

DocumentStore

Protocol for raw document text storage (pre-chunking).

class DocumentStore(Protocol):
    def add_document(
        self, doc_id: str, content: str, metadata: dict[str, Any] | None = None
    ) -> None: ...
    def get_document(self, doc_id: str) -> str | None: ...
    def list_documents(self) -> list[str]: ...
    def delete_document(self, doc_id: str) -> bool: ...

Method	Description
`add_document(doc_id, content, metadata)`	Store a document.
`get_document(doc_id)`	Retrieve text by ID. Returns `None` if not found.
`list_documents()`	Return all document IDs.
`delete_document(doc_id)`	Remove by ID. Returns `True` if found.

MemoryEntryStore

Protocol for persistent MemoryEntry storage.

class MemoryEntryStore(Protocol):
    def add(self, entry: MemoryEntry) -> None: ...
    def search(self, query: str, top_k: int = 5) -> list[MemoryEntry]: ...
    def list_all(self) -> list[MemoryEntry]: ...
    def delete(self, entry_id: str) -> bool: ...
    def clear(self) -> None: ...

Method	Description
`add(entry)`	Persist a memory entry. Overwrites if same `id` exists.
`search(query, top_k)`	Search by query string. Matching semantics are implementation-defined.
`list_all()`	Return all non-expired entries.
`delete(entry_id)`	Remove by ID. Returns `True` if found.
`clear()`	Remove all entries.

GarbageCollectableStore

Extended protocol for stores that support garbage collection of expired entries.

class GarbageCollectableStore(Protocol):
    def list_all_unfiltered(self) -> list[MemoryEntry]: ...
    def delete(self, entry_id: str) -> bool: ...

Method	Description
`list_all_unfiltered()`	Return all entries including expired ones.
`delete(entry_id)`	Remove by ID. Returns `True` if found.

In-Memory Implementations

InMemoryVectorStore

Brute-force cosine similarity vector store for development and testing.

class InMemoryVectorStore:
    def __init__(self) -> None: ...

Implements VectorStore. Thread-safe via threading.Lock.

Methods:

Method	Signature	Description
`add_embedding`	`(item_id: str, embedding: list[float], metadata: dict[str, Any] \| None = None) -> None`	Store an embedding vector with optional metadata.
`search`	`(query_embedding: list[float], top_k: int = 10) -> list[tuple[str, float]]`	Brute-force cosine similarity search. Logs a warning above 5,000 embeddings.
`delete`	`(item_id: str) -> bool`	Remove an embedding. Returns `True` if found.

[!CAUTION] Linear scan complexity O(N) per query. Use a dedicated vector database for production workloads.

InMemoryContextStore

Dict-backed context store. Thread-safe.

class InMemoryContextStore:
    def __init__(self) -> None: ...

Implements ContextStore.

Methods:

Method	Signature	Description
`add`	`(item: ContextItem) -> None`	Store a context item.
`get`	`(item_id: str) -> ContextItem \| None`	Retrieve by ID.
`get_all`	`() -> list[ContextItem]`	Return all items.
`delete`	`(item_id: str) -> bool`	Remove by ID.
`clear`	`() -> None`	Remove all items.

InMemoryDocumentStore

Dict-backed document store. Thread-safe.

class InMemoryDocumentStore:
    def __init__(self) -> None: ...

Implements DocumentStore.

Methods:

Method	Signature	Description
`add_document`	`(doc_id: str, content: str, metadata: dict[str, Any] \| None = None) -> None`	Store a document.
`get_document`	`(doc_id: str) -> str \| None`	Retrieve by ID.
`list_documents`	`() -> list[str]`	Return all document IDs.
`delete_document`	`(doc_id: str) -> bool`	Remove by ID.

InMemoryEntryStore

In-memory MemoryEntry store with filtering and search.

class InMemoryEntryStore:
    def __init__(self) -> None: ...

Implements MemoryEntryStore and GarbageCollectableStore. Thread-safe.

Methods:

Method	Signature	Description
`add`	`(entry: MemoryEntry) -> None`	Store a memory entry.
`delete`	`(entry_id: str) -> bool`	Remove by ID.
`clear`	`() -> None`	Remove all entries.
`search`	`(query: str, top_k: int = 5) -> list[MemoryEntry]`	Substring search, excluding expired entries.
`list_all`	`() -> list[MemoryEntry]`	Return all non-expired entries.
`list_all_unfiltered`	`() -> list[MemoryEntry]`	Return all entries including expired.
`get`	`(entry_id: str) -> MemoryEntry \| None`	Retrieve a single entry by ID.
`search_filtered`	`(query, top_k, *, user_id, session_id, memory_type, tags, created_after, created_before) -> list[MemoryEntry]`	Filtered search with multiple criteria.
`delete_by_user`	`(user_id: str) -> int`	Delete all entries for a user. Returns count deleted.

Persistent Implementations

JsonFileMemoryStore

JSON-file-backed persistent MemoryEntry store. Uses atomic writes (temp file + rename) to prevent corruption.

class JsonFileMemoryStore:
    def __init__(
        self,
        file_path: str | Path,
        *,
        auto_save: bool = True,
    ) -> None: ...

Parameter	Type	Default	Description
`file_path`	`str \| Path`	--	Path to the JSON file. Created on first save. Loaded on init if exists.
`auto_save`	`bool`	`True`	Persist after every mutation (`add`, `delete`, `clear`).

Implements MemoryEntryStore and GarbageCollectableStore. Thread-safe.

Methods:

Method	Signature	Description
`add`	`(entry: MemoryEntry) -> None`	Store and optionally persist to disk.
`delete`	`(entry_id: str) -> bool`	Remove by ID and optionally persist.
`clear`	`() -> None`	Remove all entries and persist.
`save`	`() -> None`	Explicitly flush to disk (atomic write).
`load`	`() -> None`	Reload entries from the JSON file.
`search`	`(query: str, top_k: int = 5) -> list[MemoryEntry]`	Substring search, excluding expired.
`list_all`	`() -> list[MemoryEntry]`	Return non-expired entries.
`list_all_unfiltered`	`() -> list[MemoryEntry]`	Return all entries including expired.
`get`	`(entry_id: str) -> MemoryEntry \| None`	Retrieve a single entry.
`search_filtered`	`(query, top_k, *, user_id, session_id, memory_type, tags, created_after, created_before) -> list[MemoryEntry]`	Filtered search.
`delete_by_user`	`(user_id: str) -> int`	Delete all entries for a user.
`export_user_entries`	`(user_id: str) -> list[MemoryEntry]`	Export all entries for a user (GDPR data portability).

[!NOTE] Not suitable for concurrent multi-process access. For multi-worker deployments, implement MemoryEntryStore with a database backend.

Storage API Reference

On this page