Implementation Plan: [Feature Name]

发布时间:2026/9/13 19:53:02
Implementation Plan: [Feature Name] Implementation Plan: [Feature Name]【免费下载链接】comet-llmDebug, evaluate, and monitor your LLM applications, RAG systems, and agentic workflows with comprehensive tracing, automated evaluations, and production-ready dashboards.项目地址: https://gitcode.com/GitHub_Trending/co/comet-llmOverview[1-2 sentence summary]Requirements[Specific requirement][Specific requirement]Affected ComponentsComponentFilesType of ChangeBackendpath/to/file.javaAdd request validation for endpointFrontendpath/to/component.tsxNew componentSDKpath/to/module.pyNew methodImplementation PhasesPhase 1: [Name]Goal: [What this achieves]Estimated complexity: Low/Medium/High[Task]-file/path[Specific change][Specific change]Test: [How to verify][Task]-file/path[Specific change]Test: [How to verify]Checkpoint: [How to verify phase is complete]Phase 2: [Name]...Testing StrategyUnit tests: [What to test]Integration tests: [What to test]Manual verification: [Steps]RisksRiskLikelihoodImpactMitigation[Risk]Low/Med/HighLow/Med/High[How to address]Open Questions[Question needing answer before proceeding]Definition of Done[Criterion for completion][Criterion for completion]模板设计要点解读 - **Affected Components 表格**按 Backend / Frontend / SDK 维度组织Type of Change 精确到为端点新增请求校验新建组件新增方法这类具体动作——与仓库的目录结构天然对齐Java 代码位于 apps/opik-backend/src/main/javaReact 组件位于 apps/opik-frontend/srcSDK 位于 sdks/python 与 sdks/typescript - **Implementation Phases**按阶段Phase推进每个阶段标注目标、复杂度评级Low/Medium/High与 **Checkpoint阶段验收点**保证每阶段结束时都有可验证的完成信号每个任务附带独立的 Test 验证方式 - **Risks 表**用 可能性×影响Low/Med/High 缓解措施 的四列结构管理不确定性 - **Definition of Done**用复选框清单定义完成标准避免感觉做完了式交付。 ## 规划原则五条质量标尺 配置文末给出五条规划原则是评审任何一份实施计划的质量标尺 - **Specific具体** —— 包含精确的文件路径与函数名 - **Testable可测试** —— 每一步都可独立验证 - **Ordered有序** —— 依赖关系清晰步骤排序正确 - **Incremental增量** —— 每个阶段都能交付价值 - **Complete完整** —— 不留任何含糊之处。 这五条原则在仓库配套的命令中得到了延续。例如 [.agents/commands/comet/work-on-jira-ticket.md](https://link.gitcode.com/i/fc8f078226d0af1398231d993330c535) 这类工作流命令要求 Agent 遵循既有技能与规划结果逐步实施[.agents/agents/meta-auditor.md](https://link.gitcode.com/i/d5a061c5a2903454f85b650bca004d1e) 中 Agents without example triggers / Agents with vague descriptions 被列为反模式正是对 Specific / Complete 原则在配置层面的呼应。 ## 配置同步机制Planner 如何进入 Cursor 与 Claude planner.md 存放在 [.agents/agents/](https://link.gitcode.com/i/3fda50bdfbbdb88f4399c11c29bc71aa)是唯一需要编辑的源文件。要将它接入不同编辑器需借助仓库根目录 [Makefile](https://link.gitcode.com/i/ac09ec26f3d001dbe81c124833643554) 提供的同步命令对应脚本实现在 [scripts/convert-frontmatter.sh](https://link.gitcode.com/i/9e6512fcf79e9793bc11282c80450a29) 与 [scripts/sync-codex.sh](https://link.gitcode.com/i/9b143cf79628b971224a8c8d2ed13393) bash make cursor # 创建 .cursor - .agents 符号链接Cursor 直接复用同一份配置 make claude # 将 .agents/ 同步到 .claude/ 并生成根目录 .mcp.json make codex # 创建 .codex 符号链接并将 rules 转为 AGENTS.override.md make clean-agents # 清理生成物保留 .agents/ 源与 .claude/ 中的本地个性化文件【免费下载链接】comet-llmDebug, evaluate, and monitor your LLM applications, RAG systems, and agentic workflows with comprehensive tracing, automated evaluations, and production-ready dashboards.项目地址: https://gitcode.com/GitHub_Trending/co/comet-llm创作声明:本文部分内容由AI辅助生成(AIGC),仅供参考