🧠 开发MCP服务器——模型上下文平台v1.0
45个工具·10个专业代理·10个预置团队·梦想家·压实机·快速工程·插件·Cron·工作树 一个AI系统可以理解你的代码库,从每次交互中学习,在你睡觉的时候做梦,并协调专业代理团队——所有这些都基于你的 *实际的* 代码,而不是通用知识。
______________________________________________________________________
为什么存在
每个开发团队都要支付一笔无形的税:
- 调试您没有编写的代码,没有上下文
- 初级开发人员等待高级开发人员在代码库的某个地方找到答案
- 为特定的系统级问题提供通用答案的人工智能
- 存在于某人脑海中的知识,当他们离开时就会死去
根本原因不是糟糕的代码。这是分散的背景。
Dev MCP Server通过为您的AI提供对实际系统的永久、不断增长的知识来解决这个问题,然后随着时间的推移使其自动变得更智能。
______________________________________________________________________
建筑
┌─────────────────────────────────────────────────────────────────────────────┐
│ Dev MCP Server v1.0 │
│ │
│ ┌─────────────────────────────────────────────────────────────────────┐ │
│ │ INGESTION LAYER │ │
│ │ Files/Dirs/Raw → FileParser → Chunker → Store → TF-IDF Indexer │ │
│ │ FileWatcher (live re-ingest on change) │ │
│ └─────────────────────────────────────────────────────────────────────┘ │
│ ↓ │
│ ┌─────────────────────────────────────────────────────────────────────┐ │
│ │ CONTEXT ENGINEERING LAYER │ │
│ │ ContextEngineer (budget/priority/rank) + Compactor (sliding window) │ │
│ │ MemoryManager (persistent facts) + Improver (usage feedback) │ │
│ └─────────────────────────────────────────────────────────────────────┘ │
│ ↓ │
│ ┌─────────────────────────────────────────────────────────────────────┐ │
│ │ 45-TOOL REGISTRY │ │
│ │ bash · file_read/write/edit/delete · grep · git_* · http_request │ │
│ │ json_query/transform · regex_test · crypto_hash · datetime │ │
│ │ run_tests · lint · api_test · mock_generate · code_complexity │ │
│ │ docker · system_info · log_analyze · dependency_analysis │ │
│ │ generate_diagram · kb_search · memory_search · think · + 20 more │ │
│ └─────────────────────────────────────────────────────────────────────┘ │
│ ↓ │
│ ┌─────────────────────────────────────────────────────────────────────┐ │
│ │ AGENT LAYER │ │
│ │ │ │
│ │ ┌─────────────────────────────────────────────────────────────┐ │ │
│ │ │ 10 Specialist Agents (real tool-use loop) │ │ │
│ │ │ Debug · Arch · Security · Docs · Refactor · Perf │ │ │
│ │ │ Test · DevOps · Data · Planner │ │ │
│ │ └─────────────────────────────────────────────────────────────┘ │ │
│ │ ┌─────────────────────────────────────────────────────────────┐ │ │
│ │ │ ORCHESTRATION │ │ │
│ │ │ TaskDecomposer → subtasks → specialist routing │ │ │
│ │ │ TeamCoordinator → 10 named teams → consolidated report │ │ │
│ │ │ PipelineEngine → 6 multi-step resilient workflows │ │ │
│ │ │ Inter-Agent Messaging → agents send messages to each other │ │ │
│ │ └─────────────────────────────────────────────────────────────┘ │ │
│ └─────────────────────────────────────────────────────────────────────┘ │
│ ↓ │
│ ┌─────────────────────────────────────────────────────────────────────┐ │
│ │ BACKGROUND INTELLIGENCE │ │
│ │ Dreamer (5 phases) · Improver · Cron Scheduler │ │
│ │ Proactive Monitor (5 checks) · Team Memory Sync │ │
│ └─────────────────────────────────────────────────────────────────────┘ │
│ ↓ │
│ ┌──────────────┐ ┌──────────────┐ ┌────────────┐ ┌───────────────────┐ │
│ │ REST API │ │ CLI (REPL) │ │ Dashboard │ │ Plugin System │ │
│ │ 40+ routes │ │ /commands │ │ /dashboard │ │ user extensions │ │
│ └──────────────┘ └──────────────┘ └────────────┘ └───────────────────┘ │
└─────────────────────────────────────────────────────────────────────────────┘______________________________________________________________________
快速开始
选项A--通过npx(无需安装)
立即运行服务器或查询代码库,而无需克隆仓库。
# In your project root (where your .env lives):
npx dev-mcp-server ingest ./src
npx dev-mcp-server query "Why is getUserById throwing?"
npx dev-mcp-server query -i # Start the interactive REPL
npx dev-mcp-server start # WebUI dashboard注: npx将在您运行命令的目录中查找.env, 因此,在运行之前,请确保您的凭据已存在。
______________________________________________________________________
选项B-本地安装(全平台)
建议使用仪表板、专业代理和Dreamer后台任务。
git clone
cd dev-mcp-server
npm install
cp .env.example .env # Add your LLM provider details
# Ingest your codebase
node cli.js ingest ./src
# Start interactive REPL (recommended)
node cli.js query -i
# Or start the REST API + Dashboard
npm start
open http://localhost:3000/dashboard______________________________________________________________________
45工具注册表
每个工具都有一个兼容的 tool_use 模式。代理在真正的API循环中以本机方式调用它们。
| 组 | 工具 |
|---|---|
| 执行 | bash |
| 文件 | file_read, file_write, file_edit (str_replace), file_delete, dir_list |
| 搜索 | grep (ripgrep+回退), find_files |
| 版本控制系统 | git_status, git_diff, git_log, git_commit (AI消息), git_branches |
| 网络 | http_request, network_check (ping/端口/dns) |
| 数据 | json_query, json_transform, regex_test, crypto_hash, datetime, schema_validate, text_diff |
| 配置 | env_read (掩盖秘密) |
| 包裹 | npm_info (列表/过时/审计) |
| 测试 | run_tests (杰斯特/摩卡), lint (埃斯林特), api_test (卷曲), mock_generate |
| 代码质量 | format_code (预处理), code_complexity |
| 分析 | log_analyze, dependency_analysis (循环/未使用/漏洞) |
| 文档 | generate_diagram (美人鱼), generate_changelog (来自git) |
| 基础设施 | docker (ps/logs/exec/inspect) |
| 系统 | system_info, process_info |
| 人工智能 | token_count, think (CoT草稿栏), run_skill |
| 知识 | kb_search, memory_search |
| 任务 | task_manage |
| 导航 | symbol_navigate (定义/参考文献/大纲) |
| 控制 | sleep |
# Execute any tool directly
curl -X POST http://localhost:3000/api/registry/execute \
-d '{"tool": "code_complexity", "input": {"path": "./src/services"}}'
# List all tools
curl http://localhost:3000/api/registry______________________________________________________________________
10名专业代理
每个代理都运行 真实工具使用循环 --它可以多次调用任何分配的工具,链接工具调用,并在步骤之间进行推理。
| 代理 | 可用工具 | 专业 |
|---|---|---|
DebugAgent | bash、fileread、grep、kb_search、git_diff、loganalyze、think。.. | 根本原因分析,精确修复 |
ArchitectureAgent | file_read、dir_list、grep、generate_schema、代码复杂性、依赖性分析。.. | 模块耦合、设计审查 |
SecurityAgent | bash、grep、apitest、envread、regextest、loganalyze。.. | OWASP审计、漏洞扫描 |
DocumentationAgent | 文件读取、kb_search、符号导航、生成图表、生成更改日志。.. | JSDoc、自述文件、API文档 |
RefactorAgent | file_read、file_edit、grep、代码复杂性、lint、format_code、text_diff… | 重复性、可读性 |
PerformanceAgent | file_read、grep、log_analyze、system_info、run_tests、bash。.. | N+1个查询,内存泄漏 |
TestAgent | file_read、file_write、run_tests、grep、mock_generate、symbol_navigation。.. | 测试写作和覆盖率 |
DevOpsAgent | bash、docker、system_info、process_info、network_check、envread。.. | CI/CD、Docker、infra |
DataAgent | jsonquery、jsontransform、schemavalidate、grep、regextest、mock_generate。.. | 模式、数据管道 |
PlannerAgent | kb_search、memory-search、任务管理、git_log、生成图表、思考。.. | 任务分解、计划 |
# Run any agent
curl -X POST http://localhost:3000/api/agents/SecurityAgent \
-d '{"task": "Audit the auth module for vulnerabilities"}'
# Decompose a complex task across multiple agents
curl -X POST http://localhost:3000/api/agents/decompose/run \
-d '{"task": "Audit payment module for security issues, perf problems, and missing docs", "parallel": false}'______________________________________________________________________
10个预先构建的团队
团队奔跑 依次 --每个代理都看到前面的发现并在此基础上构建。
| 团队 | 代理 | 用例 |
|---|---|---|
full-audit | 安全+性能+重构+文档 | 完成代码库健康检查 |
feature-review | Arch+安全+测试+文档 | 合并前的审查 |
bug-triage | 调试+性能+文档 | 对生产错误进行分类 |
onboarding | Arch+Docs+Debug+Planner | 新开发人员快速上手 |
refactor-safe | Arch+重构+测试+安全 | 安全、经过验证的重构 |
release-prep | 文档+安全+测试+DevOps | 预发布清单 |
data-audit | 数据+安全+Perf | 数据层审计 |
ci-setup | DevOps+测试+安全 | 修复或设置CI/CD |
post-mortem | 调试+性能+计划+文档 | 事件事后分析 |
greenfield | Arch+Security+Planner+Test | 规划新功能/服务 |
# Run a named team
curl -X POST http://localhost:3000/api/agents/teams/post-mortem \
-d '{"task": "Production outage last night — getUserById returning null"}'
# Auto-select the best team
curl -X POST http://localhost:3000/api/agents/teams/auto \
-d '{"task": "Prepare the v1.0 release"}'______________________________________________________________________
6条预制管道
多步骤工作流程,每一步都为下一步提供信息。弹性——失败的一步不会阻止管道。
| 管道 | 步骤 | 用例 |
|---|---|---|
debug-pipeline | 检索→ 调试→ 创建任务→ 节省内存 | 调试+操作项 |
security-audit-pipeline | 检索→ 建筑→ 安全→ 创建任务 | 完全安全扫描 |
onboarding-pipeline | 检索→ 建筑→ docs → grep待办事项 | 新开发人员入职培训 |
feature-planning-pipeline | 检索→ 分解→ plan → 创建任务 | 规划功能 |
code-review-pipeline | 检索→ git评论→ 安全→ 节省内存 | 代码审查 |
impact-analysis-pipeline | 检索→ 建筑→ 调试→ plan → 任务 | 更改影响 |
curl -X POST http://localhost:3000/api/pipelines/security-audit-pipeline \
-d '{"task": "Audit the API layer"}'______________________________________________________________________
梦想家——五阶段背景智能
每30分钟自动运行一次。不需要命令。
| 阶段 | 它做什么 |
|---|---|
| 巩固 | 合并重复的记忆,解决矛盾 |
| 模式 | 扫描代码库中的约定、反模式 |
| 建议 | 产生无人要求的主动改进 |
| 知识图谱 | 将记忆中的相关事实联系起来 |
| 修剪 | 删除超过30天的陈旧/未使用的内存 |
# Check what the dreamer found
curl http://localhost:3000/api/agents/dreamer/status
# Trigger immediately
curl -X POST http://localhost:3000/api/agents/dreamer/now______________________________________________________________________
上下文压缩程序
滑动窗口压缩使对话保持新鲜,而不会达到令牌限制。
- 窗口:逐字保存最后6条消息(始终保持最新)
- 身体:旧邮件→ 压缩摘要
- 多层:如果摘要本身太长,请再次压缩它
- 重要性评分:工具结果和关键决策权重较高
# Check if a conversation needs compaction
curl -X POST http://localhost:3000/api/compact/check \
-d '{"messages": [...]}'
# Compact
curl -X POST http://localhost:3000/api/compact/compact \
-d '{"messages": [...], "sessionId": "my-session"}'
# Deep compact (compress the summary too)
curl -X POST http://localhost:3000/api/compact/deep-compact \
-d '{"messages": [...]}'______________________________________________________________________
快速工程系统
8个内置模板+分析/改进/生成/A-B测试任何提示。
| 模板 | 说明 |
|---|---|
chain-of-thought | 回答前一步一步推理 |
role-expert | 拥有20多年经验的领域专家框架 |
few-shot | 在实际任务之前提供示例 |
structured-output | 使用模式强制JSON输出 |
critique-and-revise | 自我批评并改进答案 |
least-to-most | 先分解成更简单的子问题 |
react-agent | ReAct(Reason+Act)提示模式 |
socratic | 通过提问获得答案 |
# Analyse a prompt for weaknesses
curl -X POST http://localhost:3000/api/prompts/analyse \
-d '{"prompt": "Tell me about authentication"}'
# Auto-improve a prompt
curl -X POST http://localhost:3000/api/prompts/improve \
-d '{"prompt": "Tell me about authentication", "goal": "security audit context"}'
# Apply a template
curl -X POST http://localhost:3000/api/prompts/apply \
-d '{"template": "chain-of-thought", "variables": {"prompt": "Why is getUserById slow?"}}'
# A/B test two prompts
curl -X POST http://localhost:3000/api/prompts/ab-test \
-d '{"promptA": "...", "promptB": "...", "testInput": "..."}'
# Inject chain-of-thought into any prompt
curl -X POST http://localhost:3000/api/prompts/cot \
-d '{"prompt": "Debug this error", "style": "detailed"}'______________________________________________________________________
8种内置技能
命名的、可重用的提示工作流。用一个命令运行。
| 技能 | 它做什么 |
|---|---|
add-error-handling | 用适当的日志记录将异步代码包装在try/catch中 |
document-function | 用示例生成完整的JSDoc |
check-security | OWASP风格的目标安全审计 |
explain-flow | 跟踪函数的完整执行流 |
find-similar | 在代码库中查找重复/相似的逻辑 |
write-tests | 使用模拟生成Jest单元测试 |
performance-audit | 查找N+1个查询、内存泄漏、阻塞I/O |
migration-plan | 逐步安全迁移计划 |
node cli.js query -i
❯ /skill run write-tests getUserById
❯ /skill run check-security AuthController.js
❯ /skill run add-error-handling UserService.js______________________________________________________________________
插件系统
使用自定义插件扩展系统——添加工具、代理、路由或启动钩子。
// plugins/my-plugin.js
module.exports = {
name: 'my-plugin',
version: '1.0.0',
async register(app, { toolRegistry, agentRegistry }) {
// Add a custom tool
toolRegistry.addTool({
schema: { name: 'my_tool', description: '...', input_schema: { ... } },
execute: async (input) => 'result',
});
// Add a custom route
app.get('/api/my-plugin/hello', (req, res) => res.json({ hello: 'world' }));
},
};# List installed plugins
curl http://localhost:3000/api/plugins
# Enable/disable a plugin
curl -X POST http://localhost:3000/api/plugins/my-plugin/enable______________________________________________________________________
工作树管理器
Git工作树隔离,用于安全实验。在树枝上运行代理,不要碰你的工作树。
# Create an isolated worktree for a risky refactor
curl -X POST http://localhost:3000/api/worktrees \
-d '{"branch": "refactor/auth-system", "path": "/tmp/worktrees/auth"}'
# Run an agent in the worktree (safe, isolated)
curl -X POST http://localhost:3000/api/agents/RefactorAgent \
-d '{"task": "Refactor auth module", "cwd": "/tmp/worktrees/auth"}'
# Delete when done
curl -X DELETE http://localhost:3000/api/worktrees/auth______________________________________________________________________
Cron调度器
安排代理任务自动运行——每日安全扫描、每周审计等。
# Schedule a daily security check at 9am
curl -X POST http://localhost:3000/api/cron \
-d '{"name": "daily-security", "schedule": "0 9 * * *", "task": "Security audit of auth module", "agent": "SecurityAgent"}'
# List all scheduled jobs
curl http://localhost:3000/api/cron
# Trigger manually
curl -X POST http://localhost:3000/api/cron/daily-security/run______________________________________________________________________
团队内存同步
共享内存存储,团队中的所有代理都可以读取和写入,这些知识在会话和用户之间持续存在。
# Write to team memory
curl -X POST http://localhost:3000/api/team-memory \
-d '{"team": "backend", "content": "UserService.getUserById has a known ClassCastException bug", "type": "bug"}'
# All agents automatically read team memory on queries
# Or read it explicitly
curl "http://localhost:3000/api/team-memory?team=backend"______________________________________________________________________
代理间消息传递
代理可以在复杂的工作流中相互发送结构化消息。
# Send a message from one agent to another
curl -X POST http://localhost:3000/api/messages \
-d '{"from": "DebugAgent", "to": "SecurityAgent", "content": "Found ClassCastException — check if it exposes user data", "priority": "high"}'
# An agent reads its inbox
curl "http://localhost:3000/api/messages/inbox/SecurityAgent"______________________________________________________________________
上下文可视化工具
查看任何查询的上下文窗口中的确切内容——选择了哪些块、它们的分数、使用的令牌预算。
curl -X POST http://localhost:3000/api/context/visualize \
-d '{"query": "Why is getUserById failing?", "mode": "debug"}'
# Returns: { chunks, budgetUsed, dropped, engineeredScores, memoryContext }______________________________________________________________________
3核心查询(原始概念)
# 🐛 Why is this failing?
node cli.js query "Why is ClassCastException happening in UserService?"
# 🔍 Where is this used?
node cli.js query "Where is getUserById called across the codebase?"
# 💥 What breaks if I change this?
node cli.js query "If I change the UserDTO schema, what breaks?"______________________________________________________________________
CLI命令
node cli.js ingest
# Ingest a file or directory
node cli.js query -i # Interactive REPL (recommended)
node cli.js query "question" # Single query
node cli.js debug "error message" # Quick debug
node cli.js plan "task" # Generate execution plan
node cli.js git
node cli.js grep
# Search codebase
node cli.js tasks [list|add|done]
node cli.js memory [list|add|stats|clear]
node cli.js sessions [list|export]
node cli.js doctor # Environment health check
node cli.js stats # System overview
node cli.js cost # Token usage & cost
node cli.js clear # Reset knowledge baseREPL斜线命令
/plan Generate execution plan
/git Git status
/git commit AI-generated commit message
/git review AI code review
/agent Run a specialist agent
/team Run an agent team
/skill run
/grep
Search codebase
/lsp Go to definition
/memory Show memories
/memory add Add memory manually
/compact Compress conversation history
/cost Token usage & cost
/doctor Environment health
/help Show all commands
/exit Save & exit______________________________________________________________________
完整的REST API
# Core
GET /health # System health + KB stats
GET /dashboard # Web UI
GET / # API map
# Ingest
POST /api/ingest/file # { filePath }
POST /api/ingest/directory # { dirPath }
POST /api/ingest/raw # { content, kind, label }
DELETE /api/ingest/clear
# Query (auto-detects debug/usage/impact mode)
POST /api/query # { question, mode?, topK? }
POST /api/query/debug # { error, stackTrace? }
POST /api/query/usage # { symbol }
POST /api/query/impact # { target, changeDescription? }
POST /api/query/stream # SSE streaming
# Multi-turn Conversation
POST /api/chat # { message, convId? }
POST /api/chat/follow-up # { message, convId }
GET /api/chat # list conversations
DELETE /api/chat/:id
# Knowledge Base
GET /api/knowledge/stats
GET /api/knowledge/search?q=
GET /api/knowledge/files
POST /api/knowledge/rebuild
# Tools Registry
GET /api/registry # list all 45 tools
GET /api/registry/groups
POST /api/registry/execute # { tool, input }
# Agents (10 specialists)
GET /api/agents # list agents + stats
POST /api/agents/:name # { task }
POST /api/agents/:name/reset
POST /api/agents/decompose/run # { task, parallel? }
POST /api/agents/decompose/plan
# Teams (10 teams)
GET /api/agents/teams/list
POST /api/agents/teams/:name # { task }
POST /api/agents/teams/auto # { task }
POST /api/agents/teams/custom # { agents[], task }
# Dreamer
GET /api/agents/dreamer/status
POST /api/agents/dreamer/now
POST /api/agents/dreamer/start # { intervalMinutes? }
POST /api/agents/dreamer/stop
# Improver
GET /api/agents/improver/summary
POST /api/agents/improver/feedback # { queryId, rating, comment? }
# Pipelines
GET /api/pipelines
POST /api/pipelines/:name # { task }
POST /api/pipelines/custom/run # { steps[], task }
# Skills
GET /api/skills
POST /api/skills/:name # { target }
POST /api/skills # create custom skill
DELETE /api/skills/:name
# Prompt Engineering
GET /api/prompts/templates
POST /api/prompts/analyse # { prompt }
POST /api/prompts/improve # { prompt, goal? }
POST /api/prompts/generate # { task }
POST /api/prompts/ab-test # { promptA, promptB, testInput }
POST /api/prompts/apply # { template, variables }
POST /api/prompts/cot # { prompt, style? }
# Context
POST /api/compact/compact # { messages[] }
POST /api/compact/deep-compact
POST /api/compact/check # { messages[] }
POST /api/context/visualize # { query, mode? }
# Git
GET /api/git/status
GET /api/git/diff
POST /api/git/commit
POST /api/git/review
GET /api/git/log
GET /api/git/branches
# LSP / Symbol Navigation
GET /api/lsp/definition/:symbol
GET /api/lsp/references/:symbol
POST /api/lsp/hover # { symbol, filePath? }
GET /api/lsp/outline?path=
GET /api/lsp/symbols?q=
POST /api/lsp/rename # { oldName, newName }
# File Operations
GET /api/files/read?path=
POST /api/files/str-replace # { filePath, old_str, new_str }
POST /api/files/insert-after # { filePath, afterStr, insertText }
POST /api/files/rewrite # { filePath, content }
POST /api/files/ai-edit # { filePath, instruction }
POST /api/files/undo # { filePath }
# Shell
POST /api/tools/bash # { command, approved? }
POST /api/tools/bash/permit # { command, level }
GET /api/tools/grep?pattern=
GET /api/tools/grep/definitions/:sym
GET /api/tools/grep/imports/:mod
GET /api/tools/grep/todos
# Memory
GET /api/memory
POST /api/memory # { content, type, tags? }
DELETE /api/memory/:id
DELETE /api/memory
# Team Memory
GET /api/team-memory?team=
POST /api/team-memory # { team, content, type }
DELETE /api/team-memory/:id
# Tasks
GET /api/tasks
POST /api/tasks # { title, priority }
PATCH /api/tasks/:id
POST /api/tasks/:id/notes
DELETE /api/tasks/:id
# Sessions
GET /api/sessions
POST /api/sessions
GET /api/sessions/:id
GET /api/sessions/:id/export
DELETE /api/sessions/:id
# Planner
POST /api/plan # { task }
POST /api/plan/compact # { messages[] }
GET /api/plan/doctor
GET /api/cost?sessionId=
# Monitor
GET /api/monitor/status
GET /api/monitor/alerts
POST /api/monitor/run-all
POST /api/monitor/run/:checkId
POST /api/monitor/alerts/:id/acknowledge
POST /api/monitor/acknowledge-all
# Plugins
GET /api/plugins
POST /api/plugins/:name/enable
POST /api/plugins/:name/disable
# Worktrees
GET /api/worktrees
POST /api/worktrees # { branch, path }
DELETE /api/worktrees/:name
# Cron
GET /api/cron
POST /api/cron # { name, schedule, task, agent }
POST /api/cron/:name/run
DELETE /api/cron/:name
# Inter-Agent Messaging
POST /api/messages # { from, to, content, priority }
GET /api/messages/inbox/:agent
POST /api/messages/:id/read
# Watcher
GET /api/watcher/status
POST /api/watcher/watch # { path }
POST /api/watcher/unwatch # { path }______________________________________________________________________
关键设计原则
- AI检索,从不猜测 --每个答案都基于你的实际代码
- 背景是一个预算 --仅在最高信号信息上花费代币
- 代理商专业 --10名专家击败1名通才
- 团队按顺序思考 --每个代理都基于之前的发现
- 管道具有弹性 --失败的步骤不会终止工作流
- 梦想巩固知识 --无提示的5阶段改进
- 记忆是神经系统 --每一次交互都使系统更智能
- 反馈关闭循环 --改进者学习什么有效,什么无效
- 工具一流 --45种打字工具
tool_use整合 - 提示是经过精心设计的 --分析、改进、模板化和A/B测试每个提示
______________________________________________________________________
更多环境变量
PORT=3000 # Server port (default 3000)
ENABLE_DREAMER=true # Auto-start dreamer (default true)
DREAM_INTERVAL_MINUTES=30 # Dream frequency (default 30)
ENABLE_MONITOR=true # Auto-start monitor (default true)
LOG_LEVEL=info # Logging level______________________________________________________________________
支持摄入的文件类型
.js .ts .jsx .tsx .py .java .go .rb .php .cs .rs .cpp .c --代码 .json .yaml .yml .env .toml .xml --配置 .md .txt --文件 .log --错误日志 .sql .graphql .gql --架构 .sh .bash --脚本
