feat(ar-runtime): Grok 并行领取就绪面，引擎逻辑不变 - #7
Open
johndpope wants to merge 4 commits into
Open
Conversation
Add Grok-native agents and skills under .grok/, MCP config for both repo-root and ar-runtime cwd, an engine-driven ar-coordinator workflow as the ralph-loop stand-in, and ar-experiment-matrix for parallel seeds. Claude Code copies in ar-runtime/.claude/ stay in place.
Replace the serial next-prompt loop and "do not spawn" leaf prompts with waves of parallel engine claim workers over the ready-front. Subcoder and experiment-matrix panels follow the same fan-out (up to 32). The parent workflow/session is what scales; children stay leaves because Grok cannot nest subagents.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
这个 PR 做什么
给
ar-runtime加一条 Grok Build 执行入口：同一套 workflow engine / MCP producer / 收据与两阶段实验协议，用 claim 池把就绪面一次性铺开，提高吞吐。Claude Code 路径和引擎状态机都不改。为什么
Grok 的
agent()/parallel()会阻塞到子任务结束，不必再走 ralph-loop「每轮一个 unit」。原先 Grok 侧如果串行next-prompt，就绪面宽度 ≥ 2 时会白白排队。本 PR 让 coordinator 调已经存在的ready+claim（ar-workflow-engine.py里自组织抢占协议），按就绪面宽度 fan-out，不新增完成条件、不改complete/after-*/execute-run的语义。保证（给审查者）：
review.md/critic.md的 producer 边界。ar-runtime/.claude/；Grok 只新增.grok/与ar-runtime/.grok/config.toml。怎么验证的
新增
tests/test_grok_runtime_port.py（4 个用例）：Grok skills/agents/workflows 存在、coordinator 走ready/claim/parallel且不含next-prompt、reviewer/critic 无 Write、MCP 配置指向现有 bun server。Grok workflow 对
ar-coordinator、ar-experiment-matrix做了 canned-host 路径冒烟（编译 + 代表 args 走通 init→空 ready-front→close）。没有在本机跑完整 GPU 实验；引擎契约测试仍以ar-runtime/.claude/为 Claude 入口。范围自查
分支相对
EvoMap/main还带着 fork 上已有的 Claude skill 英文化提交（en updates/ok）。那些是译文，不是引擎行为；Grok 吞吐相关的是后两个 commit。未勾选「每个 commit 单独过测试」，是因为英文化可能对仍按中文字符串断言的 Claude 契约测试不绿。未勾选「实际运行」，是因为只做了 canned-host 冒烟，没有端到端实验。关联
无