chore(release): v0.2.7 — session tool suite, CDP Fetch interception, engine fixes

READMEs: tool count 18→23, document session_viewport/wait/screenshot/
console/storage in the capabilities list.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: root

Cargo.lock

@@ -1,6 +1,6 @@
 [package]
 name = "aginxbrowser"
-version = "0.2.6"
+version = "0.2.7"
 edition = "2021"
 authors = ["OpenCarrier Team"]
 description = "Pure server-side browser engine with built-in V8. HTTP API for scraping, search, and interactive browsing. No Chromium required."

Cargo.toml

@@ -49,7 +49,7 @@ Most new "agent browsers" are stateless, fingerprint-less one-shot renderers —
 
 - **🔐 Real TLS fingerprints** — stealth mode replicates the complete Chrome145 / Firefox133 / Safari / Edge TLS handshakes via BoringSSL (not just a UA string), switchable per request; Cloudflare Turnstile challenges wait automatically for `cf_clearance`. Fingerprint-less engines eat 403s — we get through.
 - **🤝 Stateful interactive sessions** — persistent sessions (8-minute idle keep-alive), login state injectable and exportable (`session_create(cookies=...)` ↔ `session_cookies`), surviving pagination and multi-step flows. One-shot engines throw state away.
-- **🔌 MCP native** — 18 tools as first-class citizens (not a CDP shim). Claude Code / Cursor / Claude Desktop connect in one line. HTTP + MCP dual protocol — plus a CDP bridge, so the DevTools ecosystem works too.
+- **🔌 MCP native** — 23 tools as first-class citizens (not a CDP shim). Claude Code / Cursor / Claude Desktop connect in one line. HTTP + MCP dual protocol — plus a CDP bridge, so the DevTools ecosystem works too.
 
 > Reference point: Cloudflare's Kitesurf explicitly ships neither real TLS-fingerprint negotiation nor persistent auth sessions — anti-bot and login territory is exactly where AginxBrowser plays.
 
@@ -71,7 +71,7 @@ The [local cache](#capabilities) builds on the same idea: search hits come back
 - **Tiered rendering**: static pages over plain HTTP (~100ms); V8 spins up only when JS rendering is needed (~1-2s) — 90% of the [bench](bench/README.md) page set served without spinning up V8 at all; every response reports which tier served it (`tier` field)
 - **Multi-engine meta-search**: general web (Baidu / Bing / Sogou / WeChat / Google / DuckDuckGo), news (Bing News), code (Stack Overflow, GitHub), packages (npm, PyPI), academic (arXiv), AI models (Hugging Face) — 15 engines across 7 categories, queried concurrently, merged and deduplicated. Operators can plug a private Meilisearch index into the same `/search`. Search → read in one step
 - **Image search**: `categories=images` hits Baidu/Bing image indexes and returns direct binary `image_url` links (downloadable straight to jpg/png) plus `source_url` provenance
-- **Interactive sessions**: persistent browser sessions with indexed interaction (`state/click/input/scroll/eval`) — agents browse like humans do, and `session_export` turns what an agent figured out into a runnable curl replay script (zero model tokens on re-run)
+- **Interactive sessions**: persistent browser sessions with indexed interaction (`state/click/input/scroll/eval`) — agents browse like humans do, and `session_export` turns what an agent figured out into a runnable curl replay script (zero model tokens on re-run). Session tools also cover the acting part: `session_viewport` simulates device viewports (media queries respond), `session_wait` blocks on a selector or predicate with a timeout, `session_screenshot` renders the live state, `session_console` replays the page's console ring, and `session_storage` exports/restores cookies plus localStorage for login hand-off
 - **Playback-link sniffer**: `session_network(filter=media)` extracts the m3u8/mp4/dash URLs a page's player *actually requested* at runtime — links found only in page HTML are often decoys, so the request log is the source of truth. `GET /session/{id}/har` exports the same traffic as HAR 1.2 (retained bodies included)
 - **CDP bridge**: `/json/version` + `/devtools/{kind}/{id}` WebSocket — `chromium.connectOverCDP()` from Playwright, Puppeteer, or browser-use attaches with one line ([integration guide](docs/integrations.md)). DevTools ecosystem compatibility without becoming a CDP shim
 - **File download**: streaming to disk (no memory buffering), SHA-256 integrity, resume of interrupted transfers — for binaries, archives, datasets
@@ -80,7 +80,7 @@ The [local cache](#capabilities) builds on the same idea: search hits come back
 - **JS data extraction**: `js_extract` pulls `window.__INITIAL_STATE__` and other structured data out of SPAs
 - **Screenshot rendering**: `/screenshot` endpoint (opt-in `--features screenshot`) paints the JS-rendered DOM with the diting rendering engine — pure CPU, no Chromium — to PNG. Vision input for agents
 - **TLS fingerprint spoofing**: stealth mode impersonates Chrome145/Firefox133/Safari/Edge, switchable per request
-- **MCP server**: `--mcp` mode exposes 18 tools (fetch/eval/search/download/cache + session + screenshot tools) — Claude Code / Claude Desktop / Cursor call them directly
+- **MCP server**: `--mcp` mode exposes 23 tools (fetch/eval/search/download/cache + session + screenshot tools) — Claude Code / Claude Desktop / Cursor call them directly
 - **Firecrawl compatible**: `/v1/scrape` endpoint — existing Firecrawl clients migrate by changing the base URL
 - **DNS rebinding protection**: built-in SSRF guard + post-resolution IP validation
 
@@ -203,7 +203,7 @@ aginxbrowser/
     ├── main.rs              # HTTP service entry & routing
     ├── server.rs            # Business layer (fetch/click/eval/search)
     ├── session.rs           # Interactive browser sessions
-    ├── mcp.rs               # MCP server (18 tools)
+    ├── mcp.rs               # MCP server (23 tools)
     ├── render.rs            # Tiered rendering (HTTP direct → diting browser engine)
     ├── store.rs             # Local fetch/search cache (SQLite FTS5, drift hashes)
     ├── download.rs          # Streaming file download (sha256, resume)

README.md

@@ -49,7 +49,7 @@ Agent 用浏览器要的是五件事：**看得见、读得懂、找得到、操
 
 - **🔐 真实 TLS 指纹** — stealth 模式用 BoringSSL 复刻 Chrome145 / Firefox133 / Safari / Edge 的完整 TLS 握手（不是只改 UA），可按请求切换；Cloudflare Turnstile 挑战页自动等 `cf_clearance`。无指纹引擎碰反爬就是 403，我们穿过去。
 - **🤝 有状态交互 Session** — 持久化浏览器会话（8 分钟空闲保活），登录态可注入可导出（`session_create(cookies=...)` ↔ `session_cookies`），跨翻页、跨多步流程不断。一次性引擎抓完即弃，做不了「登录 → 操作 → 再操作」。
-- **🔌 MCP 原生** — 18 个工具是一等公民（不是 CDP 套壳），Claude Code / Cursor / Claude Desktop 一行接入。HTTP + MCP 双协议之外还有 CDP 桥，DevTools 生态照样能用。
+- **🔌 MCP 原生** — 23 个工具是一等公民（不是 CDP 套壳），Claude Code / Cursor / Claude Desktop 一行接入。HTTP + MCP 双协议之外还有 CDP 桥，DevTools 生态照样能用。
 
 > 参照：Cloudflare 的 Kitesurf 明确不做真实 TLS 指纹协商、不做持久认证会话——反爬与登录正是 AginxBrowser 的地盘。
 
@@ -71,7 +71,7 @@ Agent 是照着浏览器说的话行事的，所以响应里要写清楚实际
 - **分层渲染**：静态页面纯 HTTP 直取（~100ms），需要 JS 渲染才启动 V8（~1-2s）——[bench](bench/README.md) 页面集里 90% 根本不用拉起 V8；每次响应带 `tier` 字段说明走的哪层
 - **多引擎聚合搜索**：通用网页（百度/Bing/搜狗/搜狗微信/Google/DuckDuckGo）、新闻（Bing News）、代码（Stack Overflow/GitHub）、包（npm/PyPI）、学术（arXiv）、AI 模型（Hugging Face）——15 引擎 7 分类，并发查询、合并去重；运维还可把私有 Meilisearch 索引接入同一 `/search`。Agent 一步完成"搜→读"
 - **图片搜索**：`categories=images` 接百度图片/必应图片，返回 `image_url` 二进制直链（可直接下成 jpg/png）+ `source_url` 溯源
-- **交互式 Session**：持久化浏览器会话，索引化交互（state/click/input/scroll/eval），Agent 像人一样浏览；`session_export` 把 Agent 摸索出来的操作导出成能直接跑的 curl 回放脚本，重放零模型 token
+- **交互式 Session**：持久化浏览器会话，索引化交互（state/click/input/scroll/eval），Agent 像人一样浏览；`session_export` 把 Agent 摸索出来的操作导出成能直接跑的 curl 回放脚本，重放零模型 token。操作面也补齐了：`session_viewport` 模拟设备视口（media query 会响应）、`session_wait` 按 selector/谓词带超时等待、`session_screenshot` 截会话当前状态、`session_console` 回放页面 console 环形日志、`session_storage` 导出/恢复 cookie + localStorage 方便交接登录态
 - **播放链接嗅探**：`session_network(filter=media)` 从页面播放器运行时真正发出的请求里挖 m3u8/mp4/dash 链接——写在 HTML 里的播放地址多半是诱饵，请求日志才是真相。`GET /session/{id}/har` 把同一份流量导出成 HAR 1.2（含已保留的响应体）
 - **CDP 桥**：`/json/version` + `/devtools/{kind}/{id}` WebSocket——Playwright / Puppeteer / browser-use 的 `chromium.connectOverCDP()` 一行接入（[集成指南](docs/integrations.md)）。兼容 DevTools 生态，但自己不做 CDP 套壳
 - **文件下载**：流式落盘（不吃内存）、SHA-256 校验、断点续传——二进制、压缩包、数据集用这个
@@ -80,7 +80,7 @@ Agent 是照着浏览器说的话行事的，所以响应里要写清楚实际
 - **JS 数据提取**：`js_extract` 参数，从 SPA 提 `window.__INITIAL_STATE__` 等结构化数据
 - **截图渲染**：`/screenshot` 端点（`--features screenshot`），JS 渲染后的 DOM 用自有的 diting 引擎出 PNG——纯 CPU，无 Chromium，agent 的视觉输入
 - **TLS 指纹伪装**：stealth 模式模拟 Chrome145/Firefox133/Safari/Edge，可按请求切换
-- **MCP Server**：`--mcp` 模式暴露 18 个工具（fetch/eval/click/search/download/cache + session + 截图工具），Claude Code / Claude Desktop / Cursor 直接调用
+- **MCP Server**：`--mcp` 模式暴露 23 个工具（fetch/eval/click/search/download/cache + session + 截图工具），Claude Code / Claude Desktop / Cursor 直接调用
 - **Firecrawl 兼容**：`/v1/scrape` 端点，现有 Firecrawl 客户端改 base URL 即可迁移
 - **DNS 重绑定防护**：内置 SSRF 防护 + 解析后 IP 校验
 
@@ -203,7 +203,7 @@ aginxbrowser/
     ├── main.rs              # HTTP 服务入口与路由
     ├── server.rs            # 业务层（fetch/click/eval/search）
     ├── session.rs           # 交互式浏览器会话
-    ├── mcp.rs               # MCP Server（18 个工具）
+    ├── mcp.rs               # MCP Server（23 个工具）
     ├── render.rs            # 分层渲染（HTTP 直取 → diting 浏览器引擎）
     ├── store.rs             # 本地 fetch/搜索缓存（SQLite FTS5、漂移哈希）
     ├── download.rs          # 流式文件下载（sha256、断点续传）
@@ -299,7 +299,7 @@ cargo build --release --features stealth,screenshot
 
 包含：
 - 全部 25 个 HTTP 端点（`/fetch`、`/search`、`/screenshot`、`/download`、`/v1/scrape`、`/doctor`、11 个 session 端点、CDP 发现、MCP 传输）
-- MCP Server 的 18 个工具及参数
+- MCP Server 的 23 个工具及参数
 - Claude Code / Claude Desktop / Cursor 客户端配置
 - 环境变量、错误码、站点抓取示例