bytedance/deer-flow · warning · NotImplementedError

reload_memory not supported by {type(self).__name__}

Error message

reload_memory not supported by {type(self).__name__}

What it means

MemoryManager.reload_memory raises NotImplementedError as the base-class default for backends without a cached memory document. The method's contract is 'drop the cached document and reload from storage', which only makes sense for backends that cache (DeerMem overrides it); remote backends have no local cache to invalidate, so they inherit the raising default. The docstring notes callers are expected to fall back to get_memory() when the backend does not support reload.

Source

Thrown at backend/packages/harness/deerflow/agents/memory/manager.py:395

          * ``None``  -- this backend has nothing to warm (the default). The
            host logs a "skipping" message instead of the misleading "warmed
            successfully", so a non-DeerMem backend doesn't claim a tiktoken
            cache it never touched.

        Backends with heavy one-time init override and return ``True``/``False``.
        """
        return None

    def reload_memory(
        self,
        *,
        user_id: str | None = None,
        agent_name: str | None = None,
    ) -> dict[str, Any]:
        """Drop the cached memory document and reload from storage. Default:
        unsupported (callers fall back to :meth:`get_memory`). Backends with a
        cache override."""
        raise NotImplementedError(f"reload_memory not supported by {type(self).__name__}")

    def create_fact(
        self,
        content: str,
        category: str = "context",
        confidence: float = 0.5,
        *,
        agent_name: str | None = None,
        user_id: str | None = None,
    ) -> tuple[dict[str, Any], str | None]:
        """Manually add one fact. Returns ``(memory_data, fact_id)`` -- ``fact_id``
        is None when a storage cap evicted the just-added fact. Default: unsupported."""
        raise NotImplementedError(f"create_fact not supported by {type(self).__name__}")

    def delete_fact(
        self,
        fact_id: str,
        *,

View on GitHub (pinned to 1dd6ba1acb)

Solutions

1. Call get_memory() instead — the documented fallback for backends without reload support; it returns the current view without a cache drop.
2. If you need true reload semantics (out-of-band file edits), run the deermem backend (memory.manager_class: deermem).
3. Backend authors: override reload_memory() on the MemoryManager subclass when your backend caches documents.
4. API/tooling authors: catch NotImplementedError on reload and degrade to get_memory() rather than surfacing a 500.

Example fix

# before
memory = client.reload_memory()  # or manager.reload_memory(user_id="u1")

# after
try:
    memory = manager.reload_memory(user_id="u1")
except NotImplementedError:
    memory = manager.get_memory(user_id="u1")  # documented fallback

from deerflow.agents.memory.manager import MemoryManager

if type(manager).reload_memory is MemoryManager.reload_memory:
    memory = manager.get_memory(user_id=uid)  # documented fallback path
else:
    memory = manager.reload_memory(user_id=uid)

def supports_reload(manager: MemoryManager) -> bool:
    """True when the backend has a cache and overrides reload_memory."""
    return type(manager).reload_memory is not MemoryManager.reload_memory

try:
    memory = manager.reload_memory(user_id=uid)
except NotImplementedError:
    memory = manager.get_memory(user_id=uid)  # callers fall back per the contract docstring

When it happens

Trigger: Invoking MemoryManager.reload_memory(user_id=..., agent_name=...) — e.g. POST /api/memory/reload, DeerFlowClient.reload_memory(), or a direct call — while the configured backend does not override it (noop, mem0, honcho, openviking). Out-of-band edits to DeerMem's Markdown fact files require reload(), which is why this path exists at all.

Common situations: POST /api/memory/reload after switching memory.manager_class to honcho or mem0 in config.yaml; calling DeerFlowClient.reload_memory() in an embedded integration against a non-DeerMem backend; test suites exercising the reload route against the noop backend.

Related errors

• import_memory not supported by {type(self).__name__}
• create_fact not supported by {type(self).__name__}
• delete_fact not supported by {type(self).__name__}
• update_fact not supported by {type(self).__name__}
• search not supported by {type(self).__name__}

AI-assisted analysis of bytedance/deer-flow@1dd6ba1acb (2026-08-14). Data as JSON: /api/errors/278afbae4d3aa715. Report an issue: GitHub.