LLM, Prompt completion step

Lets run the LLM.

Python
from relevanceai.steps import PromptCompletion

llm = PromptCompletion(
  prompt="Hello world",
  model="openai-gpt35"
)
llm.run()

Lets use a different model. You can get the list of supported models here https://sdk.relevanceai.com/creating-chains/supported-models.

Python
llm = PromptCompletion(
  prompt="Hello world",
  model="anthropic-claude-instant-v1"
)
llm.run()

RunChain Class

The RunChain class is a subclass of StepBase and represents a step in the relevance.ai pipeline. It is used to execute a chain of steps specified by a chain_id and params.

Constructor

Parameters

  • chain_id (str): The ID of the chain to be executed.
  • params (dict): A dictionary containing the parameters for the chain.
  • step_name (str, optional): The name of the step. Defaults to “run_chain”.
  • *args and **kwargs: Additional arguments and keyword arguments.

Properties

  • steps: Returns a list containing a single dictionary representing the step configuration.

Step Configuration

The RunChain class generates a step configuration dictionary that can be used in the relevance.ai pipeline. The step configuration contains the following fields:

  • transformation (str): The transformation type, which is set to “run_chain” for this step.
  • name (str): The name of the step, as specified in the step_name parameter.
  • foreach (str): An empty string, indicating that the step does not iterate over any input.
  • output (dict): A dictionary mapping output names to their corresponding values. The output names are generated based on the _outputs list.
  • params (dict): A dictionary containing the chain_id and params for the chain.

Output Names

The RunChain class defines a list of output names in the _outputs attribute. These names include:

  • output: The output of the chain execution.
  • state: The state of the chain execution.
  • status: The status of the chain execution.
  • errors: Any errors that occurred during the chain execution.
  • cost: The cost of the chain execution.
  • credits_used: The amount of credits used for the chain execution.
  • executionTime: The time taken for the chain execution.

The output names are used to generate the outputs property, which provides a list of fully-qualified output paths for the step configuration.

Note: The double curly braces ({{ }}) in the output configuration indicate placeholders that will be replaced with the actual values during the pipeline execution.

This RunChain class is designed to be used as a step in the relevance.ai pipeline to execute a chain of steps and retrieve the results.