diff --git a/README.md b/README.md index 0d9f15d..427758c 100644 --- a/README.md +++ b/README.md @@ -17,10 +17,10 @@ Plan, search, implement, test, review, and deliver changes through a streaming w

- Watch CODING complete a repository task and prepare a pull request + Preview: CODING searches a repository, edits code, runs checks, and prepares a pull request

-

▶ Watch the 89-second coding workflow demo

+

▶ Animated preview · Open the full 89-second video

## What is CODING? @@ -42,7 +42,8 @@ CODING is a web-based AI coding agent for real Git repositories. It turns a requ Solid borders show the current application and Eval paths. The dashed area is the documented scale-out target; it is not the default runtime topology. -

CODING architecture

+

Animated CODING architecture: request routing, agent execution, isolated evaluation, and evidence flow

+

Animated flow · Open the scalable SVG diagram

The deployment design describes **3 Agent Workers and 2 ASGI nodes**, with estimates of about **170 peak concurrent requests** and **20–30 concurrent coding/review tasks**. Treat these as design capacity estimates: this repository does not include a load-test report validating those figures. The Redis queue, repository lock, and event stream are shown as target components, not current application services. diff --git a/README.zh-CN.md b/README.zh-CN.md index 6ba3b93..6358952 100644 --- a/README.zh-CN.md +++ b/README.zh-CN.md @@ -17,10 +17,10 @@

- 点击观看 CODING 完成仓库任务并准备 Pull Request 的演示 + 预览:CODING 检索仓库、修改代码、运行检查并准备 Pull Request

-

▶ 观看 89 秒编码工作流演示

+

▶ 动态预览 · 打开 89 秒完整演示视频

## CODING 是什么 @@ -42,7 +42,8 @@ CODING 是一个面向真实 Git 仓库的 AI 编码工作台。它把用户需 实线表示当前应用和 Eval 的运行路径;虚线区域表示文档中的横向扩容目标,并非默认运行拓扑。 -

CODING 项目架构图

+

CODING 动态架构图:需求路由、Agent 执行、隔离评测与报告反馈

+

动态流程图 · 打开可缩放 SVG 源图

部署设计描述了 **3 个 Agent Worker 和 2 个 ASGI 节点**,并给出约 **170 峰值并发请求、20–30 个 coding/review 并发任务**的容量估算。仓库目前没有包含验证这些数字的压测报告,因此应视为设计容量估算。图中的 Redis 队列、仓库锁和事件流是目标组件,不是当前应用服务。 diff --git a/assets/architecture-en.svg b/assets/architecture-en.svg index afd2752..1b0e0f0 100644 --- a/assets/architecture-en.svg +++ b/assets/architecture-en.svg @@ -1,88 +1,88 @@ - - CODING system architecture - Current application, agent execution, persistence, sandbox evaluation, and a dashed scale-out deployment target. - - - - - - - - - CODING · System Architecture - Repository-aware agent · persistent sessions · isolated, reproducible evaluation - - Current path - - Target + +CODING · SYSTEM ARCHITECTUREA repository-aware coding agent with a reviewable engineering loop + + + + + + + + + +CODING · SYSTEM ARCHITECTURE +A repository-aware coding agent with a reviewable engineering loop +REQUEST → ROUTE → REASON → BUILD → VERIFY → REPORT - - 01 · PRODUCT RUNTIME - - Vue 3 workspaceTask input · plans · diffsLive status and history - - FastAPI applicationHTTP APIs · approval actionsSSE event endpoint - - Multi-layer intent routerLLM classification + rule fallbackPlanning · coding · review · QA - - Primary Agent · DeepAgents / LangGraphCoordinates task execution, delegates work, and summarizes resultsCheckpoint-backed runs · user approval and resume - - - - Dynamic agent assemblyPrompt · tools · permissionsSkills and task context - - Specialist subagentsTask breakdown · analysis · ReviewerAction decisions and findings return - - Tools and safety middlewarePath / input checks · tool recoveryModel, tool-call, and time limits - - Streaming and stateBuffered events → stable SSE protocolPostgreSQL / SQLite checkpoint + task state - - - - - Plan → implement → review → fix → verify → explain the result - - - 02 · REPOSITORY, EXECUTION AND PERSISTENCE - - Code intelligenceCodeGraph + grep hybridRepository and PR contextTargeted retrieval for agent tools - - Controlled workspacePath-scoped file backendGit branch and repository toolsRepository-scoped working copy - - OpenSandbox · Eval pathIsolated command executionSnapshot upload · patch returnResource cleanup after evaluation - - SQLite / PostgreSQLBusiness state · checkpointsWorker leases · audit eventsSQLite local default · PostgreSQL option - - Repository memoryProject structure · test commandsDecisions · history · PR contextRetrieved across sessions - - - - - 03 · REPRODUCIBLE AGENT EVAL - Pinned inputsAgent SHA · task · target SHA - Isolated case copyFresh workspace per case - Real Agent runModel · tools · retrieval traces - Acceptance gatesTarget · regression · oracle - Reviewable reportPatch · tokens · latency · HTML - - - - SCALE-OUT TARGET · DOCUMENTED DESIGN, NOT THE DEFAULT RUNTIME - 2 ASGI nodesHTTP / SSE ingress - - RedisQueue · repo lock · event stream - - 3 Agent WorkersShared PostgreSQL + workspace - Stated capacity estimate: ~170 peak requests,~20–30 concurrent coding/review tasks. - No included load-test report validates these figures. - + +APP RUNTIME + +Developer / User +Task · approval · review + +Web UI · FastAPI +REST + SSE event stream + +Intent router +LLM classification + rule fallback + +Primary Agent +Plan · delegate · code · summarize + +Code intelligence & tools +CodeGraph · grep · Git · terminal · tests + +Repository workspace +Guarded file access · reviewable diffs + +Specialist SubAgents +Analysis · read-only review · focused tasks + + + + + + +delegate / review +read / edit / test +retrieve context + +AGENT EVAL · REPRODUCIBLE CODING BENCHMARK + +Eval runner +Fixed task · agent revision · limits + +OpenSandbox +Isolated candidate repository + +Scorers & checks +Target · regression · hidden oracle + +Report & evidence +Patch · events · tokens · latency · HTML + + + + + + +candidate patch +run isolated tests +scores + traces +reproducible evidence +evaluation feedback + + +SQLite / PostgreSQL +Tasks · messages · events · checkpoints + +Repository memory +Project facts · decisions · session resume + +Scale-out design +Redis queue · worker pool · repo lock + + + +persist & resume +Dashed = documented deployment target; capacity numbers are estimates, not load-test results. + \ No newline at end of file diff --git a/assets/architecture-flow-en.gif b/assets/architecture-flow-en.gif new file mode 100644 index 0000000..9c11d0a Binary files /dev/null and b/assets/architecture-flow-en.gif differ diff --git a/assets/architecture-flow-zh.gif b/assets/architecture-flow-zh.gif new file mode 100644 index 0000000..1e7d1cb Binary files /dev/null and b/assets/architecture-flow-zh.gif differ diff --git a/assets/architecture-zh.svg b/assets/architecture-zh.svg index 20955a3..458d533 100644 --- a/assets/architecture-zh.svg +++ b/assets/architecture-zh.svg @@ -1,86 +1,88 @@ - - CODING 项目系统架构 - 展示当前应用、Agent 执行、持久化、沙箱评测和虚线标注的横向扩容目标。 - - - - - - - - - CODING · 项目系统架构 - 理解仓库的 Agent · 可恢复会话 · 隔离且可复跑的评测 - 当前链路 - 扩容目标 + +CODING · 项目系统架构面向真实代码仓库,形成可审查、可复现的研发闭环 + + + + + + + + + +CODING · 项目系统架构 +面向真实代码仓库,形成可审查、可复现的研发闭环 +提交需求 → 意图路由 → Agent 执行 → 测试验证 → 评测报告 - - 01 · 产品运行链路 - - Vue 3 工作台提交任务 · 查看计划与补丁实时状态与历史会话 - - FastAPI 应用HTTP API · 审批操作SSE 事件接口 - - 多层意图路由LLM 语义分类 + 关键词规则兜底规划 · 编码 · 审查 · QA - - 主 Agent · DeepAgents / LangGraph统筹任务、委派工作并汇总结果Checkpoint 恢复 · 用户审批后继续执行 - - - - 按任务动态装配Prompt · 工具 · 权限Skills 与任务上下文 - - 专业 SubAgent任务拆解 · 分析 · Reviewer动作决策与审查发现回传 - - 工具与安全中间件路径/参数校验 · 工具异常恢复模型、工具调用和时长上限 - - 流式输出与运行状态队列缓冲 → 稳定 SSE 事件协议PostgreSQL / SQLite checkpoint 与任务状态 - - - - - 规划 → 编码 → 审查 → 修复 → 验证 → 汇报结果 - - - 02 · 仓库、执行与持久化 - - 代码检索CodeGraph + grep 混合检索仓库与 PR 上下文供 Agent 工具定位相关代码 - - 受控项目工作区限制路径范围的文件后端Git 分支与仓库工具按仓库管理工作副本 - - OpenSandbox · Eval 路径隔离环境中执行命令上传源码快照 · 回传补丁评测完成后清理沙箱资源 - - SQLite / PostgreSQL业务状态 · 会话 checkpointWorker 租约 · 审计事件SQLite 本地默认 · PostgreSQL 可选 - - 仓库长期记忆项目结构 · 测试命令技术决策 · 历史 · PR 上下文跨会话检索与复用 - - - - - 03 · 可复现 Agent Eval - 固定输入版本Agent SHA · 题目 · 目标 SHA - 独立案例副本每题创建新工作区 - 启动真实 Agent模型 · 工具 · 检索轨迹 - 验收门禁目标 · 回归 · 隐藏验收 - 可复核报告补丁 · Token · 耗时 · HTML - - - - 横向扩容目标 · 部署设计,并非当前默认运行链路 - 2 个 ASGI 节点接收 HTTP / SSE - - Redis队列 · 仓库锁 · 事件流 - - 3 个 Agent Worker共享 PostgreSQL 与工作区 - 容量估算:峰值约 170 个请求,20–30 个 coding/review 并发任务。 - 仓库未附验证该数字的压测报告。 - + +应用运行链路 + +开发者 / 用户 +需求 · 审批 · 代码审查 + +Web 工作台 · FastAPI +REST 接口 + SSE 事件流 + +意图路由 +LLM 语义分类 + 规则兜底 + +主 Agent +计划 · 委派 · 编码 · 汇总 + +代码检索与工具 +CodeGraph · grep · Git · 终端 · 测试 + +仓库工作区 +受控文件访问 · 可审查 diff + +专业 SubAgent +分析 · 只读审查 · 专项子任务 + + + + + + +任务委派 / 审查 +读取 / 修改 / 测试 +检索上下文 + +AGENT EVAL · 可复现编码能力评测 + +评测运行器 +固定题目 · Agent 版本 · 资源限制 + +OpenSandbox +隔离的候选项目副本 + +评分器与验收 +目标测试 · 回归测试 · 隐藏验收 + +报告与证据 +补丁 · 事件 · Token · 耗时 · HTML + + + + + + +候选补丁 +隔离运行测试 +评分与轨迹 +可复现证据 +评测反馈 + + +SQLite / PostgreSQL +任务 · 消息 · 事件 · checkpoint + +仓库长期记忆 +项目事实 · 技术决策 · 会话恢复 + +横向扩容设计 +Redis 队列 · Worker 池 · 仓库锁 + + + +持久化与恢复 +虚线表示文档中的部署目标;并发数字为估算值,尚非压测结果。 + \ No newline at end of file diff --git a/assets/demo/coding-agent-demo-preview.gif b/assets/demo/coding-agent-demo-preview.gif new file mode 100644 index 0000000..4faac3f Binary files /dev/null and b/assets/demo/coding-agent-demo-preview.gif differ