[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-co11ap5e-coding-agent":3},{"id":4,"slug":5,"name":6,"summary":7,"stack":8,"background":13,"responsibility":14,"workflow":15,"architecture":22,"highlights":23,"retrospective":28,"repositoryUrl":29,"liveUrl":30,"featured":31},6,"co11ap5e-coding-agent","co11ap5e_coding_agent（TDD 编程助手）","基于自研 ReAct Agent 框架扩展的 AI 编程助手，内置强制 TDD 循环，自动完成写测试 → 跑测试（RED）→ 写实现 → 再跑测试（GREEN）的完整开发流程。",[9,10,11,12],"Python","tiktoken","TDD","AST","在自研 ReAct 框架之上构建一个面向真实工程的 Coding Agent：用 Repo Map 理解代码库、用 tiktoken 精确计费并自动压缩上下文、用强制 TDD 循环保证产出经过测试验证。","独立完成架构设计与实现。\n\n- 实现 TDD 红绿循环：先写测试再写实现，全部通过才收尾\n- 用 AST 提取函数 \u002F 类 \u002F 导入符号，按重要性生成 token 预算内的 Repo Map\n- 基于 tiktoken 精确统计 token，超预算自动把旧消息压缩为摘要\n- 实现 9 个内置工具：文件读写改、Shell、代码搜索、Git 操作\n- 工具参数容错解析，执行错误回填给 LLM 自行修正",[16,17,18,19,20,21],"接收编程任务","扫描代码库生成 Repo Map","按 TDD 流程先写测试并运行（RED）","再写实现并运行（GREEN）","失败则读错误信息修正后重试","全部通过后输出总结","自研 ReAct 主循环 + TDD 纪律；AST Repo Map 提供代码库概览；tiktoken 精确 token 预算与自动摘要压缩；项目记忆持久化到 .coding_agent\u002Fproject_memory.json。",[24,25,26,27],"强制 TDD 循环，测试驱动开发","AST Repo Map 在预算内理解代码库","tiktoken 精确计数与自动压缩","9 个内置工具覆盖文件 \u002F Shell \u002F 搜索 \u002F Git","与主流 Coding Agent 对比，本项目定位教学 \u002F 学习级：每一层（ReAct、工具系统、记忆、TDD）都可逐行解释，透明可控是核心价值。","https:\u002F\u002Fgithub.com\u002FS1rryNut\u002Fco11ap5e_coding_agent",null,true]