langchain-ai/langchain · error · ValueError

Did not find {key}, please add an environment variable `{key

Error message

Did not find {key}, please add an environment variable `{key}` which contains it, or pass `{key}` as a named parameter.

What it means

Default error from the `from_env` lookup closure when the environment variable is absent, no default was provided, and no custom `error_message` was set. The message names the exact variable LangChain expected and tells you the two remedies: set the env var or pass the value as a named parameter.

Source

Thrown at libs/core/langchain_core/utils/utils.py:419

            The value from the environment.
        """
        if isinstance(key, (list, tuple)):
            for k in key:
                if k in os.environ:
                    return os.environ[k]
        if isinstance(key, str) and key in os.environ:
            return os.environ[key]

        if isinstance(default, (str, type(None))):
            return default
        if error_message:
            raise ValueError(error_message)
        msg = (
            f"Did not find {key}, please add an environment variable"
            f" `{key}` which contains it, or pass"
            f" `{key}` as a named parameter."
        )
        raise ValueError(msg)

    return get_from_env_fn


@overload
def secret_from_env(key: str | Sequence[str], /) -> Callable[[], SecretStr]: ...


@overload
def secret_from_env(key: str, /, *, default: str) -> Callable[[], SecretStr]: ...


@overload
def secret_from_env(
    key: str | Sequence[str], /, *, default: None
) -> Callable[[], SecretStr | None]: ...

View on GitHub (pinned to e32fa9a52e)

Solutions

Set the named environment variable: `export <KEY>=<value>` or add `<KEY>=...` to `.env` and ensure it is loaded before construction.
Pass the parameter explicitly to the constructor: `ChatProvider(api_key=...)`.
Verify with `python -c "import os; print(os.environ.get('<KEY>'))"` that the variable is actually visible to the process.

Example fix

# before
from langchain_community.chat_models import ChatHuggingFace
model = ChatHuggingFace()  # ValueError: Did not find HUGGINGFACEHUB_API_TOKEN ...

# after
import os
os.environ["HUGGINGFACEHUB_API_TOKEN"] = "hf_..."
model = ChatHuggingFace()

Defensive patterns

Strategy: validation

Validate before calling

import os

def require_env(keys: list[str]) -> None:
    missing = [k for k in keys if k not in os.environ]
    if missing:
        raise SystemExit(
            f"Missing environment variables: {missing}. "
            "Set them or pass the corresponding named parameter."
        )

require_env(["HUGGINGFACEHUB_API_TOKEN"])

Try / catch

try:
    llm = build_llm()
except ValueError as e:
    if "Did not find" in str(e) and "environment variable" in str(e):
        raise SystemExit(f"Configuration incomplete: {e}") from e
    raise

Prevention

Verify with `os.environ.get(KEY)` in a preflight check before building clients.
In Docker/CI, assert required secrets exist in the entrypoint so failures carry deployment context.
Pass values explicitly in tests instead of relying on ambient env state.

When it happens

Trigger: Constructing an integration whose field default uses `from_env('SOME_KEY')` (no default, no custom message) while `SOME_KEY` is not in `os.environ`; passing a `list`/`tuple` of alternative keys where none are set also lands here.

Common situations: Deploying to a new environment (staging/prod/container) without propagating secrets; `.env` not loaded before import-time instantiation; shell vs subprocess env differences; variable renamed but code not updated.

Related errors

AI-assisted analysis of langchain-ai/langchain@e32fa9a52e (2026-08-14). Data as JSON: /api/errors/63ce91e21ad8d7d4. Report an issue: GitHub.