Add CogsguardTeacherPolicy for role-based vibe assignment #4983
Conversation
How to use the Graphite Merge QueueAdd either label to this PR to merge it via the merge queue:
You must have a Graphite account in order to use the merge queue. Sign up using this link. An organization admin has enabled the Graphite Merge Queue in this repository. Please do not merge from GitHub as this will restart CI on PRs being processed by the merge queue. This stack of pull requests is managed by Graphite. Learn more about stacking. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b1dbd3e278
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| available_vibes = [name[len("change_vibe_") :] for name in change_vibe_actions] | ||
| if role_vibes is None: | ||
| role_vibes = [vibe for vibe in available_vibes if vibe != "default"] | ||
| if not role_vibes: | ||
| role_vibes = available_vibes |
There was a problem hiding this comment.
When role_vibes is None, this defaults to all change_vibe_* actions except default (here), which in Cogsguard missions includes non-role vibes like heart and gear. The Nim Cogsguard policy only handles gear/miner/scout/aligner/scrambler and falls through to noop for other vibes (see step in cogsguard_agents.nim), so forcing heart at episode start yields a no-op agent for the whole episode, and forcing gear yields a random role instead of the intended deterministic rotation. Consider filtering the default list to the actual role vibes or to those with explicit behaviors.
Useful? React with 👍 / 👎.
Add CogsguardTeacherPolicy for Cogsguard Agents
This PR introduces a new
CogsguardTeacherPolicythat wraps the Cogsguard agent policy and forces agents to adopt specific vibes at the beginning of each episode. The policy cycles through available vibes for each agent, ensuring they take on different roles across episodes.Key additions:
CogsguardTeacherPolicythat delegates toCogsguardAgentsMultiPolicyafter forcing an initial vibeCogsguardAgentsMultiPolicythat uses the new Nim-basedCogsguardPolicyThe policy tracks episode state to ensure vibes are only forced at the beginning of each episode and rotates through available roles for each agent.