Skip to main content
RunArtifacts is the structured record of an agent run. Your agent populates it during execution, and verifiers consume it to evaluate task completion.

​
Fields

FieldTypeRequiredDescription
versionstrYesSchema version (currently "0.1")
task_idstrYesTask identifier
taskstrYesTask instruction text
modelstrNoModel used (e.g., "gpt-4o")
providerstrNoModel provider (e.g., "openai")
messageslist[dict]NoFull message history (role + content)
tool_callslist[ToolCall]NoAll tool invocations made
tool_resultslist[ToolCallResult]NoResults from each tool call (parallel to tool_calls)
metadatadictNoArbitrary metadata
final_observationstrNoAgent’s final output/summary
errorstrNoError message if the run failed
steps_takenintYesNumber of tool calls executed
max_stepsintNoStep limit (if configured)
created_atstrYesISO 8601 timestamp

​
Helper Methods

MethodDescription
record_message(role, content)Append a message to the history
record_tool_call(call, result)Record a tool call + result, increments steps_taken
set_final_observation(text)Set the agent’s final output
set_error(text)Record an error
to_dict()Serialize to a JSON-compatible dict
dump(path)Write artifacts to a JSON file

​
ToolCall

@dataclass
class ToolCall:
    tool_server: str          # e.g. "email"
    tool_name: str            # e.g. "send_email"
    parameters: dict[str, Any]  # tool-specific parameters

​
ToolCallResult

@dataclass
class ToolCallResult:
    observation: Any          # response from the tool server
    is_error: bool = False    # True if the tool call failed