_run_llm() — langchain Function Reference

Architecture documentation for the _run_llm() function in runner_utils.py from the langchain codebase.

Function python LangChainCore MessageInterface calls 2 called by 1

Entity Profile

LangChainCore→ MessageInterface→ _run_llm() — langchain Function Reference

Dependency Diagram

graph TD
  916eedb8_7d10_72b9_9829_105f8ada65ab["_run_llm()"]
  8253c602_7d0c_9195_a7e1_3e9b19304131["runner_utils.py"]
  916eedb8_7d10_72b9_9829_105f8ada65ab -->|defined in| 8253c602_7d0c_9195_a7e1_3e9b19304131
  385c1e91_d947_1192_8746_ee1dd66ceb54["_run_llm_or_chain()"]
  385c1e91_d947_1192_8746_ee1dd66ceb54 -->|calls| 916eedb8_7d10_72b9_9829_105f8ada65ab
  cb0a59f9_bf61_2368_e170_04f16da99179["_get_prompt()"]
  916eedb8_7d10_72b9_9829_105f8ada65ab -->|calls| cb0a59f9_bf61_2368_e170_04f16da99179
  c56a3c0a_b0e2_287c_e948_c9d9eb18b351["_get_messages()"]
  916eedb8_7d10_72b9_9829_105f8ada65ab -->|calls| c56a3c0a_b0e2_287c_e948_c9d9eb18b351
  style 916eedb8_7d10_72b9_9829_105f8ada65ab fill:#6366f1,stroke:#818cf8,color:#fff

Relationship Graph

Source Code

libs/langchain/langchain_classic/smith/evaluation/runner_utils.py lines 861–926

def _run_llm(
    llm: BaseLanguageModel,
    inputs: dict[str, Any],
    callbacks: Callbacks,
    *,
    tags: list[str] | None = None,
    input_mapper: Callable[[dict], Any] | None = None,
    metadata: dict[str, Any] | None = None,
) -> str | BaseMessage:
    """Run the language model on the example.

    Args:
        llm: The language model to run.
        inputs: The input dictionary.
        callbacks: The callbacks to use during the run.
        tags: Optional tags to add to the run.
        input_mapper: function to map to the inputs dictionary from an Example
        metadata: Optional metadata to add to the run.

    Returns:
        The LLMResult or ChatResult.

    Raises:
        ValueError: If the LLM type is unsupported.
        InputFormatError: If the input format is invalid.
    """
    # Most of this is legacy code; we could probably remove a lot of it.
    if input_mapper is not None:
        prompt_or_messages = input_mapper(inputs)
        if isinstance(prompt_or_messages, str) or (
            isinstance(prompt_or_messages, list)
            and all(isinstance(msg, BaseMessage) for msg in prompt_or_messages)
        ):
            llm_output: str | BaseMessage = llm.invoke(
                prompt_or_messages,
                config=RunnableConfig(
                    callbacks=callbacks,
                    tags=tags or [],
                    metadata=metadata or {},
                ),
            )
        else:
            msg = (
                "Input mapper returned invalid format: "
                f" {prompt_or_messages}"
                "\nExpected a single string or list of chat messages."
            )
            raise InputFormatError(msg)
    else:
        try:
            llm_prompts = _get_prompt(inputs)
            llm_output = llm.invoke(
                llm_prompts,
                config=RunnableConfig(
                    callbacks=callbacks,
                    tags=tags or [],
                    metadata=metadata or {},
                ),
            )
        except InputFormatError:
            llm_inputs = _get_messages(inputs)
            llm_output = llm.invoke(
                **llm_inputs,
                config=RunnableConfig(callbacks=callbacks, metadata=metadata or {}),
            )
    return llm_output