Skip to main content

portia.execution_agents.context

Context builder that generates contextual information for the PlanRun.

This module defines a set of functions that build various types of context required for the run execution. It takes information about inputs, outputs, clarifications, and execution metadata to build context strings used by the agent to perform tasks. The context can be extended with additional system or user-provided data.

generate_main_system_context​

def generate_main_system_context(
        system_context_extensions: list[str] | None = None) -> list[str]

Generate the main system context.

Arguments:

  • system_context_extensions list[str] | None - Optional list of strings to extend the system context.

Returns:

  • list[str] - A list of strings representing the system context.

generate_input_context​

def generate_input_context(inputs: list[Variable],
                           previous_outputs: dict[str, Output]) -> list[str]

Generate context for the inputs and indicate which ones were used.

Arguments:

  • inputs list[Variable] - The list of input variables for the current step.
  • previous_outputs dict[str, Output] - A dictionary of previous step outputs.

Returns:

  • list[str] - A list of strings representing the input context.

generate_clarification_context​

def generate_clarification_context(clarifications: ClarificationListType,
                                   step: int) -> list[str]

Generate context from clarifications for the given step.

Arguments:

  • clarifications ClarificationListType - A list of clarification objects.
  • step int - The step index for which clarifications are being generated.

Returns:

  • list[str] - A list of strings representing the clarification context.

generate_context_from_execution_context​

def generate_context_from_execution_context(
        context: ExecutionContext) -> list[str]

Generate context from the execution context.

Arguments:

  • context ExecutionContext - The execution context containing metadata and additional data.

Returns:

  • list[str] - A list of strings representing the execution context.

build_context​

def build_context(ctx: ExecutionContext, step: Step, plan_run: PlanRun) -> str

Build the context string for the agent using inputs/outputs/clarifications/ctx.

Arguments:

  • ctx ExecutionContext - The execution context containing agent and system metadata.
  • step Step - The current step in the PlanRun including inputs.
  • plan_run PlanRun - The current run containing outputs and clarifications.

Returns:

  • str - A string containing all relevant context information.