nemoguardrails._compat.langchain_kwargs

0.21 -> 0.22 LangChain config migration helper.

Detects LangChain Python-side flags in model.parameters when the default framework is active and raises a clear error at LLMRails construction, so a stale 0.21 LangChain config surfaces during init rather than as an opaque HTTP 400 deep in a guardrail call.

Remove in 0.23.0. After 0.23 any unrecognized parameter is forwarded verbatim to the OpenAI-compatible HTTP client; the wire’s HTTP 400 is the user’s signal to clean up.

Module Contents

Functions

Name	Description
`_canonical_name_for`	-
`_detect_provider_alias`	-
`_violations_for`	Return a list of (model_type, action) tuples for one model.
`check_langchain_kwargs`	Raise ValueError if any model carries LangChain Python-side flags.

Data

_LANGCHAIN_BASE_FLAGS

_PROVIDER_PREFIXED_ALIAS

API

nemoguardrails._compat.langchain_kwargs._canonical_name_for(
    matched_canonical: str
) -> str

nemoguardrails._compat.langchain_kwargs._detect_provider_alias(
    name: str
) -> typing.Optional[str]

nemoguardrails._compat.langchain_kwargs._violations_for(
    model_type: str,
    engine: str,
    parameters: dict
) -> typing.List[typing.Tuple[str, str]]

Return a list of (model_type, action) tuples for one model.

nemoguardrails._compat.langchain_kwargs.check_langchain_kwargs(
    models: typing.Iterable[nemoguardrails.rails.llm.config.Model],
    active_framework: str
) -> None

Raise ValueError if any model carries LangChain Python-side flags.

No-op when the active framework is anything other than default; LangChain-flavored kwargs are valid on the LangChain framework.

nemoguardrails._compat.langchain_kwargs._LANGCHAIN_BASE_FLAGS = frozenset({'streaming', 'disable_streaming', 'verbose', 'cache', 'callbacks', 't...

nemoguardrails._compat.langchain_kwargs._PROVIDER_PREFIXED_ALIAS = re.compile('^(?P<prefix>[a-zA-Z]\\w*?)_(?P<canonical>api_key|base_url|api_base|e...