Skip to content

simple_chain_builder

Single-step chain builder module

Attributes

LLM_METADATA_BIND_KWARGS module-attribute 

LLM_METADATA_BIND_KWARGS = 'bind_kwargs'

Classes

SimpleChainBuilder

Bases: BaseChainBuilder

Attributes

chain_param_overrides class-attribute instance-attribute 
chain_param_overrides: Dict[str, Dict] = Field(
    default_factory=dict,
    title="Chain Parameter Overrides",
    description="",
)
chain_params class-attribute instance-attribute 
chain_params: Dict[str, Any] = Field(
    ...,
    title="Chain Parameters",
    description="Contains parameters that are passed to the chain builder and are lazily evaluated when the chain is loaded. The key in this dictionary is used for matching LLM names such that specific parameter overrides can be configured per LLM.",
)
llm_ref class-attribute instance-attribute 
llm_ref: str | None = Field(
    default=None,
    title="LLM Reference",
    description="Reference to the chain LLM",
)

Functions

base_chain 
base_chain(
    llm_metadata: Dict[str, Any] | None = None
) -> Tuple[
    SimpleChainBuilderParams,
    RunnableSerializable,
    Dict[str, RunnableLambda],
]

Creates the base chain components

build 
build() -> RunnableSerializable

Builds the chain configuration.

This method is responsible for building the chain configuration based on the parameters provided in the configuration.

Returns:

  • RunnableSerializable ( RunnableSerializable ) –

    The chain configuration

derive_chain_params 
derive_chain_params(
    llm_ref: str | None,
) -> SimpleChainBuilderParams

Get the LLM parameters based on the llm_ref.

Parameters:

  • llm_ref (str | None) –

    The reference to the LLM parameters.

Returns:

Raises:

  • ValueError –

    If no LLM parameters are found for the given llm_ref and no defaults are configured on the chain.

derive_chain_token_limits 
derive_chain_token_limits(
    base_template: Dict[str, Any] | None = None,
    default_output_token_pct: float = 0.25,
) -> Tuple[int, int, BaseTokenizer]

Derives the token limits for a given chain.

Intended use case is for routines that will be stuffing data into a language model and need to know the maximum input and output token limits.

Example usage:

max_input_tokens, max_output_tokens, tokenizer = derive_chain_token_limits( chain_name=”cognitive_derive_values” )

Parameters:

  • base_template (Dict[str, Any], default: None ) –

    The base template for the prompt. Defaults to None.

  • default_output_token_pct (float, default: 0.25 ) –

    The default percentage of output tokens. Defaults to 0.25.

Returns:

  • Tuple[int, int, BaseTokenizer] –

    Tuple[int, int, BaseTokenizer]: A tuple containing the maximum input tokens, maximum output tokens, and the tokenizer.

Raises:

  • ValueError –

    If the chain factory is not found.

SimpleChainBuilderParams

Bases: BaseModel

Simple chain builder parameters object

Attributes

additional_echo_fields class-attribute instance-attribute 
additional_echo_fields: List[str] | None = Field(
    default=None,
    title="Additional Echo Fields",
    description="",
)
llm_bind_kwargs class-attribute instance-attribute 
llm_bind_kwargs: Dict[str, Any] = Field(
    default_factory=dict,
    title="LLM Bind Kwargs",
    description="",
)
mapper_to_str_kwargs class-attribute instance-attribute 
mapper_to_str_kwargs: Dict[str, Any] = Field(
    default_factory=dict,
    title="Mapper to Str Kwargs",
    description="",
)
model_config class-attribute instance-attribute 
model_config = ConfigDict(
    populate_by_name=True,
    use_enum_values=False,
    extra="forbid",
    strict=False,
    arbitrary_types_allowed=True,
)
output_parser class-attribute instance-attribute 
output_parser: Any = Field(
    default=None,
    title="Output Parser",
    description="Optional Langchain output parser",
)
response_field_name class-attribute instance-attribute 
response_field_name: str = Field(
    default=FIELD_RESPONSE,
    title="Response Field Name",
    description="Name of the field in the response that contains the output",
)
template class-attribute instance-attribute 
template: str = Field(
    ...,
    title="Template",
    description="Template to use for the chain",
)
template_format class-attribute instance-attribute 
template_format: str = Field(
    default="f-string",
    title="Template Format",
    description="",
)

Functions

validate_output_parser classmethod 
validate_output_parser(v)

Functions