Re-exports goal command components, providing a thin API to run goal-driven agent sessions.
# Source: https://github.com/OpenHands/software-agent-sdk | License: MIT | Copyright 2026 OpenHands contributors
# Provided via justclaw.it.com; keep this notice with the code.
"""The ``/goal`` command: judge-driven, self-continuing goal completion.
A conversation-level command (not a critic) that drives the agent toward an
objective: it sends the objective, runs the agent, judges completion with a
second LLM, and re-prompts until the goal is done or a cap is reached.
The decision logic lives in :class:`GoalController` (transport-agnostic, no
I/O); :func:`run_goal` is a thin synchronous driver over it. An async
agent-server task can reuse the same controller with its own I/O loop.
Usage::
from openhands.sdk.conversation.goal import run_goal
outcome = run_goal(conversation, "ma
... (truncated -- full source via MCP)
See the full source, get the GitHub permalink, and search for more like it.
Get a free API key