langchain-ai/deepagents · error · TypeError

`create_summarization_middleware` expects `model` to be a `B

Error message

`create_summarization_middleware` expects `model` to be a `BaseChatModel` instance.

What it means

`create_summarization_middleware` builds defaults from the model's profile, so it requires an actual `BaseChatModel` instance. Passing anything else (a model name string, a callable/factory, None) raises `TypeError` at the isinstance check.

Source

Thrown at libs/deepagents/deepagents/middleware/summarization.py:1691

        model: Resolved `BaseChatModel` instance.

            Use `resolve_model()` first if needed for model strings.
        backend: Backend instance for persisting conversation history.
        summary_prompt: Prompt template for generating summaries.
        trim_tokens_to_summarize: Max tokens to include when generating summary.
        token_counter: Function to count tokens in messages.

    Returns:
        Configured `SummarizationMiddleware` instance.

    Raises:
        TypeError: If `model` is not a `BaseChatModel` instance.
    """
    from langchain.chat_models import BaseChatModel as RuntimeBaseChatModel  # noqa: PLC0415

    if not isinstance(model, RuntimeBaseChatModel):
        msg = "`create_summarization_middleware` expects `model` to be a `BaseChatModel` instance."
        raise TypeError(msg)

    defaults = compute_summarization_defaults(model)
    return SummarizationMiddleware(
        model=model,
        backend=backend,
        trigger=defaults["trigger"],
        keep=defaults["keep"],
        token_counter=token_counter,
        summary_prompt=summary_prompt,
        trim_tokens_to_summarize=trim_tokens_to_summarize,
        truncate_args_settings=defaults["truncate_args_settings"],
    )


def create_summarization_tool_middleware(
    model: str | BaseChatModel,
    backend: BackendProtocol,
    *,

View on GitHub (pinned to a1af029e6e)

Solutions

Instantiate the model first, e.g. `init_chat_model("openai:gpt-4.1")` or `ChatOpenAI(model=...)`, and pass the instance
Check the caller (e.g. `create_deep_agent`) isn't forwarding a raw config string
Resolve string identifiers to instances before calling the factory

Example fix

// before
create_summarization_middleware(model="anthropic:claude-sonnet-4-5")
// after
from langchain.chat_models import init_chat_model
create_summarization_middleware(model=init_chat_model("anthropic:claude-sonnet-4-5"))

Defensive patterns

Strategy: type-guard

Validate before calling

from langchain.chat_models import BaseChatModel
if not isinstance(model, BaseChatModel):
    model = init_chat_model(model)  # resolve string ids to instances

Type guard

def is_chat_model(model: object) -> bool:
    return isinstance(model, BaseChatModel)

Try / catch

try:
    mw = create_summarization_middleware(model=model)
except TypeError as e:
    if "BaseChatModel" in str(e):
        model = init_chat_model(model)  # if model was a string id
    else:
        raise

Prevention

Pass instantiated chat models everywhere, not name strings
Centralize model construction in one factory function
Add isinstance assertions at configuration boundaries

When it happens

Trigger: Calling `create_summarization_middleware(model="openai:gpt-4.1", ...)` with a string identifier or other non-instance value instead of an instantiated chat model.

Common situations: Confusing the string-model convention accepted elsewhere (e.g. spec `model` fields) with this factory; passing a lazy model factory; refactors that replaced instances with model names.

Understand the failure class

Background: "Wrong argument type", "must be a string", "expected Array or Prism::Scope": TypeError and ArgumentError when a library receives a value of the wrong type — this error's family across 28 libraries.

Related errors

AI-assisted analysis of langchain-ai/deepagents@a1af029e6e (2026-08-29). Data as JSON: /api/errors/c1d8ec654b86407f. Report an issue: GitHub.