📦v2.0.0

v2.0.0

v2.0.0 is a major upgrade from v1.2.7, introducing Playwright support for the first time and restructuring browser-
assisted search and fetch flows.
v2.0.0 相比 v1.2.7 是一次较大的升级，首次引入了 Playwright，并重构了浏览器增强的搜索与抓取链路。

Highlights / 主要更新

• First-time Playwright integration for browser-assisted workflows
  首次引入 Playwright，用于支持浏览器增强能力
• Added flexible Playwright integration options:
  PLAYWRIGHT_PACKAGE, PLAYWRIGHT_MODULE_PATH, PLAYWRIGHT_EXECUTABLE_PATH, PLAYWRIGHT_WS_ENDPOINT,
  PLAYWRIGHT_CDP_ENDPOINT
  新增灵活的 Playwright 接入方式：
  PLAYWRIGHT_PACKAGE、PLAYWRIGHT_MODULE_PATH、PLAYWRIGHT_EXECUTABLE_PATH、PLAYWRIGHT_WS_ENDPOINT、
  PLAYWRIGHT_CDP_ENDPOINT
• Introduced SEARCH_MODE with request, auto, and playwright modes
  新增 SEARCH_MODE 配置，支持 request、auto 和 playwright 三种模式
• Refactored Bing browser fallback flow with improved request-to-browser switching and Playwright availability
  detection
  重构 Bing 的浏览器回退链路，优化 request 到浏览器模式的切换逻辑，以及 Playwright 可用性探测
• Added a browser cookie bridge so browser-acquired cookies can be reused by request-based fetches
  新增浏览器 Cookie bridge，可将浏览器获取到的 Cookie 复用到 request 请求中
• Enhanced fetchWebContent with browser cookie retry, browser HTML fallback, and retrievalMethod reporting
  增强 fetchWebContent，支持 browser cookie retry、browser HTML fallback，并新增 retrievalMethod 字段标识抓取方式
• Improved CSDN and Zhihu article fetching with browser-assisted retry and better whitespace normalization
  增强 CSDN 和知乎文章抓取，支持浏览器辅助重试，并优化正文空白清理
• Unified LinuxDo and Zhihu site search on top of the shared Bing capability
  将 LinuxDo 和知乎的站内搜索统一复用到共享的 Bing 能力之上
• Added live tests covering Bing, web content fetching, and article fetching scenarios
  新增多组 live test，覆盖 Bing、网页抓取和文章抓取场景
• Updated documentation with Playwright usage, CDP examples, and Windows PowerShell examples
  更新文档，补充了 Playwright 用法、CDP 示例以及 Windows PowerShell 示例

Breaking Changes / 破坏性变更

• SEARCH_MODE currently affects Bing only
  SEARCH_MODE 当前仅对 Bing 生效
• Browser-assisted behavior is newly introduced in this release, so users adopting Playwright-based workflows should
  review the new environment variables and runtime options
  浏览器增强能力是本版本首次引入，因此如果你要启用基于 Playwright 的工作流，需要检查新的环境变量和运行配置

Upgrade Notes / 升级说明

• If you only use request-based behavior, upgrading should be straightforward
  如果你只使用基于 request 的能力，这次升级通常可以直接完成
• If you want browser-assisted behavior, install and configure Playwright manually after upgrading
  如果你需要浏览器增强能力，请在升级后手动安装并配置 Playwright
• CDP is supported via PLAYWRIGHT_CDP_ENDPOINT, which can be used to connect to an existing local Chrome/Chromium
  session
  支持通过 PLAYWRIGHT_CDP_ENDPOINT 使用 CDP，可连接到本地已有的 Chrome/Chromium 会话

Why 2.0.0 / 为什么是 2.0.0

This release introduces a new browser-assist architecture, adds Playwright as a new optional capability, and changes
how search and content fetching can fall back from request mode to browser mode.
这次更新引入了一套新的浏览器增强架构，首次加入 Playwright 作为可选能力，并改变了搜索与内容抓取从 request 模式回退到浏
览器模式的整体方式。

What's Changed

• Feature/feature playwright by @Aas-ee in #51

Full Changelog: v1.2.7...v2.0.0

📦 v1.2.5

Release Notes (v1.2.0 -> v1.2.5) / 更新说明（v1.2.0 -> v1.2.5）

Highlights / 重点更新

• Added generic web fetch tool: fetchWebContent
  新增通用网页抓取工具：fetchWebContent
• Added engine-name normalization (e.g. Bing, DuckDuckGo, linux.do)
  新增搜索引擎名称标准化（如 Bing、DuckDuckGo、linux.do）
• Hardened web-fetch security (SSRF guard + response size limits)
  强化网页抓取安全（SSRF 防护 + 响应体大小限制）
• Fixed Vercel/Serverless runtime error: MCP error -32603: keyValidator._parse is not a function
  修复 Vercel/Serverless 运行时错误：MCP error -32603: keyValidator._parse is not a function

---

Added / 新增

• New tool: fetchWebContent for public HTTP(S) URLs, including Markdown and HTML extraction
  新增 fetchWebContent 工具，支持公开 HTTP(S) 链接抓取，含 Markdown/HTML 内容提取

• Configurable MCP tool names via env vars:
  新增工具名环境变量配置：

  • MCP_TOOL_SEARCH_NAME
  • MCP_TOOL_FETCH_LINUXDO_NAME
  • MCP_TOOL_FETCH_CSDN_NAME
  • MCP_TOOL_FETCH_GITHUB_NAME
  • MCP_TOOL_FETCH_JUEJIN_NAME
  • MCP_TOOL_FETCH_WEB_NAME

• New test coverage:
  新增测试覆盖：

  • test:web-content
  • test:web-content:live
  • test:engine-normalization
  • test:url-safety

---

Changed / 变更

• search.engines input is now normalized before enum validation
  search.engines 参数在枚举校验前先做标准化处理

• Logging output moved to stderr for better MCP/stdio compatibility
  日志输出统一到 stderr，提升 MCP/stdio 兼容性

---

Fixed / 修复

• Fixed Brave parser compatibility with updated page structure
  修复 Brave 页面结构变更导致的解析问题

• Improved LinuxDo search fallback behavior
  优化 LinuxDo 搜索回退逻辑

• Fixed schema parsing/runtime compatibility issue in serverless deployments
  修复 Serverless 部署下的 schema 解析兼容性问题

• Resolved dependency mismatch causing:
  解决依赖不一致导致的报错：

  • MCP error -32603: keyValidator._parse is not a function

---

Security / 安全

• SSRF protections for fetchWebContent:
  fetchWebContent 增加 SSRF 防护：

  • block localhost/private/link-local targets
    拦截 localhost / 私网 / 链路本地地址
  • re-validate redirect final URL
    对重定向后的最终 URL 再次校验

• Response/body size limits added before full processing
  在完整处理前增加响应体大小限制，降低内存风险

---

Compatibility / 兼容性

• No breaking changes
  无破坏性变更

• Existing canonical engine values keep working as-is
  现有标准引擎参数保持兼容（如 bing、duckduckgo）

What's Changed

• feat: Configurable tool names by @lightdot in #24
• chore: replace console.log with console.error fix #23 by @Aas-ee in #25
• fix(brave,linuxdo): update selectors for SvelteKit SSR & add Brave fallback by @Zacklinkk in #39
• chore(package): Update package name and add zod dependency by @Aas-ee in #40
• fix(docs): fix readme by @Aas-ee in #41
• Feature/feature web fetch by @Aas-ee in #42
• feat(web): Enhanced security and reliability of web content acquisition by @Aas-ee in #43

New Contributors

• @lightdot made their first contribution in #24
• @Zacklinkk made their first contribution in #39

Full Changelog: v1.2.0...v1.2.5

📦 v1.2.0

Release v1.2.0 - Add Configurable Engines and STDIO Mode

🚀 New Features

Configurable Search Engines

• New Environment Variable: ALLOWED_SEARCH_ENGINES
  • Restrict available search engines by specifying comma-separated engine names
  • Provides better control over which engines can be used in your deployment
  • If the default engine is not in the allowed list, the first allowed engine becomes the default
  • Example: ALLOWED_SEARCH_ENGINES=duckduckgo,bing,exa

Mode Switching Parameters

• Enhanced MODE Configuration: MODE environment variable now supports three options:
  • stdio - STDIO mode only (ideal for MCP client integration)
  • http - HTTP server mode only (for web-based access)
  • both - Both STDIO and HTTP modes (default behavior)
• Flexible Deployment: Choose the most appropriate mode for your use case

📋 Configuration Reference

Variable	Default	Options	Description
ALLOWED_SEARCH_ENGINES	empty (all available)	Comma-separated engine names	Limit which search engines can be used
MODE	both	both, http, stdio	Server operation mode

🔧 Usage Examples

Restricted Engine Usage

# Linux/macOS
ALLOWED_SEARCH_ENGINES=duckduckgo,bing DEFAULT_SEARCH_ENGINE=duckduckgo npx open-websearch@latest

# Windows PowerShell
$env:ALLOWED_SEARCH_ENGINES="duckduckgo,bing"; $env:DEFAULT_SEARCH_ENGINE="duckduckgo"; npx open-websearch@latest

Mode-specific Deployments

# STDIO only (for MCP clients)
MODE=stdio npx open-websearch@latest

# HTTP only (for web access)
MODE=http npx open-websearch@latest

# Windows CMD STDIO mode
set MODE=stdio && set DEFAULT_SEARCH_ENGINE=duckduckgo && npx open-websearch@latest

📦 v1.1.4.2

📦 v1.1.4.2

Release Date: August 4, 2025 | 发布日期： 2025年8月4日
Tag: v1.1.4.2
Status: Stable / Production Ready

---

🚀 New Features & Enhancements | 新功能与增强

• ✅ NPX Quick Start Support | NPX 快速启动支持
  Now supports instant deployment with npx open-websearch@latest - no installation required!
  现在支持使用 npx open-websearch@latest 即时部署 - 无需安装！

• ✅ Juejin Search & Article Fetching | 掘金搜索与文章获取
  Added support for Juejin (掘金) platform search and full article content extraction
  新增掘金平台搜索支持和完整文章内容提取功能

• ✅ GitHub README Fetching | GitHub README 获取
  New tool fetchGithubReadme for extracting README content from GitHub repositories
  新增 fetchGithubReadme 工具，用于提取 GitHub 仓库的 README 内容

• ✅ Enhanced Multi-engine Support | 增强的多引擎支持
  Now includes: Bing, DuckDuckGo, Baidu, CSDN, Exa, Brave, Juejin
  现已包含：Bing、DuckDuckGo、百度、CSDN、Exa、Brave、掘金

• ✅ Cross-platform Environment Variable Support | 跨平台环境变量支持
  Better Windows PowerShell compatibility and cross-env support for local development
  更好的 Windows PowerShell 兼容性和本地开发的 cross-env 支持

• ✅ Improved Docker Deployment | 改进的 Docker 部署
  Enhanced Docker configuration with better environment variable handling
  增强的 Docker 配置，更好的环境变量处理

---

🛠️ Setup Options | 安装选项

• NPX Quick Start | NPX 快速启动 (npx open-websearch@latest) - NEW! | 新功能！
• Global Installation | 全局安装 (npm install -g open-websearch)
• Local Development | 本地开发 (npm install && npm run build)
• Docker Deployment | Docker 部署 (docker-compose or docker run)
• Compatible with Claude Desktop, Claude VSCode extension, and Cherry Studio
  兼容 Claude Desktop、Claude VSCode 扩展和 Cherry Studio

---

🔧 Available Tools | 可用工具

• search - Multi-engine web search with configurable engines | 可配置引擎的多引擎网络搜索
• fetchCsdnArticle - Extract full CSDN article content | 提取完整 CSDN 文章内容
• fetchJuejinArticle - Extract full Juejin article content NEW! | 新功能！
• fetchGithubReadme - Fetch GitHub repository README files NEW! | 新功能！

---

📌 Configuration | 配置

Environment Variables | 环境变量:

• DEFAULT_SEARCH_ENGINE: bing, duckduckgo, exa, brave | 默认搜索引擎
• USE_PROXY / PROXY_URL: HTTP proxy support | HTTP 代理支持
• ENABLE_CORS / CORS_ORIGIN: CORS configuration | CORS 配置
• PORT: Custom server port | 自定义服务器端口

---

📌 Known Limitations | 已知限制

• Some engines may block excessive requests (rate limiting)
  某些引擎可能会阻止过多请求（频率限制）
• Parsing depends on HTML structure of target sites
  解析依赖于目标站点的 HTML 结构
• IP-based Anti-crawling & JavaScript Challenges | 基于IP的反爬限制与JavaScript挑战
  Different IP types may face varying levels of anti-crawling restrictions and challenges:
  不同IP类型可能面临不同程度的反爬限制和挑战：
  • Cloud server IPs may be more strictly limited, blocked, or face JavaScript challenges (CAPTCHA, bot detection)
    云服务器IP可能受到更严格的限制、封禁，或面临JavaScript挑战（验证码、机器人检测）
• Not all results include meta info (e.g., description may be missing)
  并非所有结果都包含元信息（例如，描述可能缺失）
• Personal use only - please comply with search engine terms of service
  仅限个人使用 - 请遵守搜索引擎服务条款

---

🔄 Migration from v1.1 | 从 v1.1 升级说明

No breaking changes - existing configurations remain compatible. Simply update to the latest version to access new features.
无破坏性更改 - 现有配置保持兼容。只需更新到最新版本即可使用新功能。

---

📦 v1.1

Release Date: July 10, 2025
Tag: v1.1

---

🚀 What's New

• ✅ Multi-engine web search support
  Engines: Bing, DuckDuckGo, Baidu, CSDN, Exa, Brave
• ✅ API key-free usage
  Fully functional without requiring authentication or keys
• ✅ Proxy support
  Allows configuration of HTTP proxy to bypass regional restrictions
• ✅ Structured output
  Includes title, URL, and description fields in search results
• ✅ Default engine configuration
• ✅ Result limit customization
• ✅ Article fetching tool for CSDN
  Enables full content extraction from individual articles

---

🛠️ Setup Options

• Local: npm install && npm run build
• Docker: docker-compose or docker run
• Compatible with: Claude Desktop, Claude VSCode Extension, Cherry Studio

---