BerriAI/litellm · error · AttributeError

LLM client cache lazy import: unknown attribute {name!r}

Error message

LLM client cache lazy import: unknown attribute {name!r}

What it means

The LLM client cache lazy-import handler accepts exactly two names: 'LLMClientCache' (the class) and 'in_memory_llm_clients_cache' (a singleton instance, created once and cached). Any other attribute routed to this handler raises AttributeError 'LLM client cache lazy import: unknown attribute <name>'. It is a closed whitelist, not a general import path.

Source

Thrown at litellm/_lazy_imports.py:403

    if name in _globals:
        return _globals[name]

    # Import the class
    module: Final = importlib.import_module("litellm.caching.llm_caching_handler")
    LLMClientCache: Final = getattr(module, "LLMClientCache")

    # If they want the class itself, return it
    if name == "LLMClientCache":
        _globals["LLMClientCache"] = LLMClientCache
        return LLMClientCache

    # If they want the singleton instance, create it (only once)
    if name == "in_memory_llm_clients_cache":
        instance: Final = LLMClientCache()
        _globals["in_memory_llm_clients_cache"] = instance
        return instance

    raise AttributeError(f"LLM client cache lazy import: unknown attribute {name!r}")


def _lazy_import_http_handlers(name: str) -> Any:
    """
    Handler for HTTP clients - has special logic for creating client instances.

    This one is different because:
    - These aren't just imports, they're actual client instances that need to be created
    - They need configuration (timeout, etc.) from the module globals
    - They use factory functions instead of direct instantiation
    """
    _globals: Final = get_litellm_globals()

    if name == "module_level_aclient":
        # Create an async HTTP client using the factory function
        from litellm.llms.custom_httpx.http_handler import get_async_httpx_client

        # Get timeout from module config (if set)

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Use the exact names: litellm.LLMClientCache for the class or litellm.in_memory_llm_clients_cache for the singleton instance.
2. Inspect the singleton's public API (dir(litellm.in_memory_llm_clients_cache)) instead of guessing sibling attributes.
3. If you need client invalidation, look for the documented reset/clear helpers on the cache object or router rather than unmapped names.
4. Search litellm/_lazy_imports.py for other whitelists if the symbol you want lives in a different category.

Example fix

# before
cache = litellm.llm_client_cache  # AttributeError: LLM client cache lazy import: unknown attribute 'llm_client_cache'

# after
cache = litellm.in_memory_llm_clients_cache  # singleton instance
cache_cls = litellm.LLMClientCache                 # the class

ALLOWED = {'LLMClientCache', 'in_memory_llm_clients_cache'}
if name not in ALLOWED:
    raise AttributeError(f'{name!r} not offered; use one of {sorted(ALLOWED)}')

import litellm

def client_cache_attr(name: str):
    if name not in ('LLMClientCache', 'in_memory_llm_clients_cache'):
        return None
    return getattr(litellm, name)

try:
    cache = getattr(litellm, 'in_memory_llm_clients_cache')
except AttributeError as e:
    if 'LLM client cache' in str(e):
        cache = None  # name not in the two-entry whitelist; check spelling
    else:
        raise

When it happens

Trigger: Trying to access litellm.in_memory_llm_clients_cache / litellm.LLMClientCache with a typo, or probing sibling names (e.g. 'llm_client_cache', 'client_cache') that are not in the two-name whitelist; introspection code enumerating attributes can also land here.

Common situations: Custom caching/teardown code that wants to flush per-model client caches; guessing attribute names instead of checking docs/source; hasattr sweeps over the module during debugging.

Related errors

• module {__name__!r} has no attribute {name!r}
• {category} lazy import: unknown attribute {name!r}
• Utils module lazy import: unknown attribute {name!r}
• HTTP handlers lazy import: unknown attribute {name!r}
• input must be a string or a list

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/8c38427724c1c801. Report an issue: GitHub.