What causes "Invalid config type for provider {provider}" in mem0ai/mem0?

Pass config as a plain dict of the provider's fields: {'provider': 'qdrant', 'config': {'host': ..., 'collection_name': ...}}.

mem0ai/mem0 · error · ValueError

Invalid config type for provider {provider}

Error message

Invalid config type for provider {provider}

What it means

ValueError from the same VectorStoreConfig validator when config is neither a dict nor an instance of the provider's pydantic config class (e.g. QdrantConfig). The validator accepts a plain dict (coerced into the config class) or an already-constructed config object of exactly the right type; anything else — including a config object built for a DIFFERENT provider — is rejected.

Source

Thrown at mem0/vector_stores/configs.py:60

    def validate_and_create_config(self) -> "VectorStoreConfig":
        provider = self.provider
        config = self.config

        if provider not in self._provider_configs:
            raise ValueError(f"Unsupported vector store provider: {provider}")

        module = __import__(
            f"mem0.configs.vector_stores.{provider}",
            fromlist=[self._provider_configs[provider]],
        )
        config_class = getattr(module, self._provider_configs[provider])

        if config is None:
            config = {}

        if not isinstance(config, dict):
            if not isinstance(config, config_class):
                raise ValueError(f"Invalid config type for provider {provider}")
            return self

        # also check if path in allowed kays for pydantic model, and whether config extra fields are allowed
        if "path" not in config and "path" in config_class.__annotations__:
            config["path"] = f"/tmp/{provider}"

        self.config = config_class(**config)
        return self

View on GitHub (pinned to 001c235229)

Solutions

Pass config as a plain dict of the provider's fields: {'provider': 'qdrant', 'config': {'host': ..., 'collection_name': ...}}.
Or pass an instance of the exact matching class, e.g. from mem0.configs.vector_stores.qdrant import QdrantConfig; QdrantConfig(...).
Keep provider and config class in one place (single source of truth) so they cannot drift apart.

Example fix

# before
MemoryConfig(vector_store=VectorStoreConfig(provider="elasticsearch", config=ChromaConfig(...)))

# after
MemoryConfig(vector_store=VectorStoreConfig(provider="elasticsearch", config={"host": "localhost", "port": 9200}))

Defensive patterns

Strategy: type-guard

Validate before calling

def build_vector_store_config(provider: str, config):
    if isinstance(config, dict):
        return VectorStoreConfig(provider=provider, config=dict(config))
    from mem0.vector_stores.configs import VectorStoreConfig
    cls_name = VectorStoreConfig._provider_configs[provider]
    import importlib
    cls = getattr(importlib.import_module(f"mem0.configs.vector_stores.{provider}"), cls_name)
    if not isinstance(config, cls):
        raise TypeError(f"config for {provider} must be a dict or {cls.__name__}")
    return VectorStoreConfig(provider=provider, config=config)

Type guard

def is_valid_provider_config(provider: str, config) -> bool:
    import importlib
    from mem0.vector_stores.configs import VectorStoreConfig
    if isinstance(config, dict):
        return True
    try:
        cls = getattr(importlib.import_module(f"mem0.configs.vector_stores.{provider}"),
                      VectorStoreConfig._provider_configs[provider])
        return isinstance(config, cls)
    except (KeyError, ModuleNotFoundError):
        return False

Prevention

Default to plain dicts for provider config; let pydantic validate field names and types.
Keep the provider string and its config object generated from one code path so they cannot diverge.
Add unit tests asserting your config dict constructs successfully for each provider you ship.

When it happens

Trigger: Passing vector_store={'provider': 'chroma', 'config': QdrantConfig(...)} (mismatched provider/config class), config='some string', config=[...], or a custom dataclass instead of the provider's config class or a dict.

Common situations: Switching the provider string in config but forgetting to change the config object; wrapping config in nested dicts of the wrong shape; passing an OSS MemoryConfig instance where a provider-specific config is expected.

Related errors

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/94823bd2ee5155b3. Report an issue: GitHub.