Disallow Any Generics

What it does

disallow_any_generics is a type-checking flag in langchain-core that enforces strict generic-type discipline. When enabled, a class that uses Any as a generic parameter, such as Runnable[dict, Any], raises a TypeError at runtime or fails Pydantic schema validation. It forces precise input/output types instead of the catch-all Any.

Why it matters

Weakly typed LLM pipelines are a common source of bugs: a ChatModel emits a BaseMessage, your chain expects a str, and you hunt an AttributeError for hours. This flag closes that gap at the source and acts as a CI guardrail for teams shipping SDKs built on LangChain.

Example

$ def my_func(x: str) -> str:
    return x.upper()
runnable = RunnableLambda(my_func)
schema = config_schema(runnable)
print("Schema fields:", list(schema.get("fields", {}).keys()))
print("Type safety: PASS")

Schema fields: ['func', 'name', 'tags', 'metadata', 'recurse', 'step']
Type safety: PASS

Common flags

Runnable[Input, Output]

TypedRunnable generic specifying exact input/output types

config_schema(runnable)

Extracts the Pydantic config schema for a runnable

disallow_any_generics

Runtime check rejecting Any as a generic type parameter