DRAFT: Critic agent that forks generator settings for PR-style review #1706
Conversation
Coverage Report •
|
||||||||||||||||||||
|
@OpenHands do a /codereview-roasted and help me address all review comments in this PR. When you are done, you should help me write a script in examples/0 folder that demonstrate this critic work in action. You can refer to the existing delagation example, as well as the example introduced in #1269 |
|
I'm on it! xingyaoww can track my progress at all-hands.dev |
Co-authored-by: openhands <openhands@all-hands.dev>
Final summary (double-checked)✅ Checklist against your request
🔍 Conciseness / extraneous changes check
📌 What changed (since the last state on main)
🚀 Push status
|
|
Looks like there are a few issues preventing this PR from being merged!
If you'd like me to help, just leave a comment, like Feel free to include any additional details that might help me get this PR into a better state. You can manage your notification settings |
There was a problem hiding this comment.
We should put it under https://github.com/OpenHands/software-agent-sdk/tree/main/examples/01_standalone_sdk
|
|
||
| if __name__ == "__main__": | ||
| raise SystemExit(main()) | ||
| """ |
There was a problem hiding this comment.
This seems not ideal. Should we make a hook ideally accepting a callback function as well?
| def _extract_llm(self, events: Sequence[LLMConvertibleEvent]) -> LLM | None: | ||
| for event in events: | ||
| agent = getattr(event, "agent", None) | ||
| llm = getattr(agent, "llm", None) | ||
| if isinstance(llm, LLM): | ||
| return llm | ||
|
|
||
| for event in events: | ||
| if not isinstance(event, SystemPromptEvent): | ||
| continue | ||
| agent = getattr(event, "agent", None) | ||
| llm = getattr(agent, "llm", None) | ||
| if isinstance(llm, LLM): | ||
| return llm | ||
|
|
||
| return None |
There was a problem hiding this comment.
We don't need this, since self.llm is already there
| `not_pass`. | ||
| """ | ||
|
|
||
| llm: LLM | None = None |
There was a problem hiding this comment.
This should be llm: LLM
It is a required arg
| def _extract_agent(self, events: Sequence[LLMConvertibleEvent]) -> Agent | None: | ||
| for event in events: | ||
| agent = getattr(event, "agent", None) | ||
| if isinstance(agent, Agent): | ||
| return agent | ||
|
|
||
| for event in events: | ||
| if not isinstance(event, SystemPromptEvent): | ||
| continue | ||
| agent = getattr(event, "agent", None) | ||
| if isinstance(agent, Agent): | ||
| return agent | ||
|
|
||
| return None |
There was a problem hiding this comment.
Similar to the get planning agent function in the openhands.tools.preset - we should reconstruct an Agent instance ourselves
Summary
AgentReviewCritic, aCriticBaseimplementation that spawns a separate OpenHands agent to review the current git diff.Fixes #1704
Checklist
@xingyaoww can click here to continue refining the PR
Agent Server images for this PR
• GHCR package: https://github.com/OpenHands/agent-sdk/pkgs/container/agent-server
Variants & Base Images
eclipse-temurin:17-jdknikolaik/python-nodejs:python3.12-nodejs22golang:1.21-bookwormPull (multi-arch manifest)
# Each variant is a multi-arch manifest supporting both amd64 and arm64 docker pull ghcr.io/openhands/agent-server:de8bb36-pythonRun
All tags pushed for this build
About Multi-Architecture Support
de8bb36-python) is a multi-arch manifest supporting both amd64 and arm64de8bb36-python-amd64) are also available if needed