Skip to content

feat: runtime/agent split + visual servoing + StepFun + bilingual README#4

Merged
ACCSCI merged 2 commits into
masterfrom
feat/runtime-agent-split
Jun 22, 2026
Merged

feat: runtime/agent split + visual servoing + StepFun + bilingual README#4
ACCSCI merged 2 commits into
masterfrom
feat/runtime-agent-split

Conversation

@ACCSCI

@ACCSCI ACCSCI commented Jun 22, 2026

Copy link
Copy Markdown
Owner

变更概述 / Summary

架构重构 / Architecture Refactor

  • Runtime/Agent 分离:lib/runtime/ (loop, events, checkpoint, verifier) + lib/agents/ (browser-agent, registry) — agents are plain data, runtime is the engine
  • 12 种独立 RuntimeEvent 类型,替代 catch-all update
  • Wall-clock timeout 改用 chrome.alarms(SW 重启后仍有效)
  • Checkpoint 持久化到 chrome.storage.session
  • Tool 错误作为 Observation(safeExecute 模式)

功能新增 / New Features

  • StepFun (阶跃星辰) 支持,reasoning 通过 think 标记透传
  • CDP 视觉瞄准闭环:aim → 对比 → cancel → re-aim
  • DPR 从截图/视口比值计算,不依赖 window.devicePixelRatio
  • Session 导出到剪贴板/文件
  • 首次响应的合成流式气泡

E2E 测试 / E2E Tests

  • 41-search-engine-cdp: 搜索引擎 CDP 交互
  • 41-stepfun-tab-task: StepFun 标签页管理
  • 42-msgstore-reconnect: SW 重启后消息存储重连
  • 更新 visual-servoing, minimax-bing-task, data-layer-routing

配置 / Config

  • 新增中英双语 README / Bilingual README
  • .gitignore 添加 research/ 排除
  • 升级 AI SDK v4→v6, Zod v3→v4

@gemini-code-assist gemini-code-assist Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request updates the README.md to provide a bilingual (Chinese and English) documentation layout, translating sections such as features, quick start, architecture, and supported LLM providers. The review feedback highlights a minor typo in the environment variables table ('阶跃星辰提供商标' instead of '阶跃星辰提供商') and identifies several incorrect or non-existent LLM model names (e.g., 'GPT-4.1', 'Claude Sonnet 4', and 'step-3.7-flash') that should be corrected to ensure accurate documentation.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

Comment thread README.md Outdated
|---|---|---|
| `MINIMAX_API_KEY` | MiniMax 提供商 / MiniMax provider | [platform.minimaxi.com](https://platform.minimaxi.com/user-center/basic-information/keys) |
| `MIMO_API_KEY` | 小米 MiMo 提供商 / Xiaomi MiMo provider | [platform.xiaomimimo.com](https://platform.xiaomimimo.com/user-center/api-keys) |
| `STEPFUN_API_KEY` | 阶跃星辰提供商标 / StepFun provider | [platform.stepfun.com](https://platform.stepfun.com/user-center/basic-information/keys) |

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

在表格中,“阶跃星辰提供商标”存在输入错误,多了一个“标”字,应为“阶跃星辰提供商”。

Suggested change
| `STEPFUN_API_KEY` | 阶跃星辰提供商标 / StepFun provider | [platform.stepfun.com](https://platform.stepfun.com/user-center/basic-information/keys) |
| `STEPFUN_API_KEY` | 阶跃星辰提供商 / StepFun provider | [platform.stepfun.com](https://platform.stepfun.com/user-center/basic-information/keys) |

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

✅ Fixed in 6b76a3d

Comment thread README.md
Comment on lines +168 to +173
| **OpenAI** | GPT-4o, GPT-4.1, o3 等 | 原生支持 `@ai-sdk/openai` |
| **Anthropic** | Claude Sonnet 4, Opus 4 等 | 原生支持 `@ai-sdk/anthropic` |
| **OpenAI 兼容** / OpenAI-compatible | 任何 OpenAI API 兼容端点 | 自定义 Base URL + Key |
| **小米 MiMo** / Xiaomi MiMo | MiMo-M3 | 支持视觉,工具调用能力强 |
| **MiniMax** | MiniMax-M3 | 支持视觉 + 推理 |
| **阶跃星辰** / StepFun | step-3.7-flash | 通过 `<think>` 标记透传推理 |

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

模型名称中存在几处拼写或版本错误:\n1. OpenAI: 不存在 GPT-4.1 模型,建议改为 GPT-4 TurboGPT-4。\n2. Anthropic: 目前最新的模型为 Claude 3.5 SonnetClaude 3 Opus,不存在 Claude Sonnet 4Opus 4。\n3. StepFun: 阶跃星辰目前没有 step-3.7-flash 模型,其主流模型为 step-1.5-flashstep-2-16k 等。\n\n建议修正这些模型名称以保证文档的准确性。

Suggested change
| **OpenAI** | GPT-4o, GPT-4.1, o3 等 | 原生支持 `@ai-sdk/openai` |
| **Anthropic** | Claude Sonnet 4, Opus 4 等 | 原生支持 `@ai-sdk/anthropic` |
| **OpenAI 兼容** / OpenAI-compatible | 任何 OpenAI API 兼容端点 | 自定义 Base URL + Key |
| **小米 MiMo** / Xiaomi MiMo | MiMo-M3 | 支持视觉,工具调用能力强 |
| **MiniMax** | MiniMax-M3 | 支持视觉 + 推理 |
| **阶跃星辰** / StepFun | step-3.7-flash | 通过 `<think>` 标记透传推理 |
| **OpenAI** | GPT-4o, GPT-4 Turbo, o3 等 | 原生支持 `@ai-sdk/openai` |\n| **Anthropic** | Claude 3.5 Sonnet, Claude 3 Opus 等 | 原生支持 `@ai-sdk/anthropic` |\n| **OpenAI 兼容** / OpenAI-compatible | 任何 OpenAI API 兼容端点 | 自定义 Base URL + Key |\n| **小米 MiMo** / Xiaomi MiMo | MiMo-M3 | 支持视觉,工具调用能力强 |\n| **MiniMax** | MiniMax-M3 | 支持视觉 + 推理 |\n| **阶跃星辰** / StepFun | step-1.5-flash | 通过 `<think>` 标记透传推理 |

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

模型名称已验证正确,无需修改:

  • GPT-4.1 — OpenAI 于 2025 年 4 月发布,现为 GPT-4 系列最新版本
  • Claude Sonnet 4 / Opus 4 — Anthropic 于 2025 年发布,当前最新模型
  • step-3.7-flash — 阶跃星辰当前主力模型,项目代码中实际使用的模型

Gemini 的训练数据截止较早,这些模型在其知识范围内确实不存在,但目前已正式发布。

@ACCSCI ACCSCI merged commit f5a57dc into master Jun 22, 2026
@ACCSCI ACCSCI deleted the feat/runtime-agent-split branch June 22, 2026 05:42
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant