dsh-vision-router 3.0.2 → 3.0.3

package/README.md

@@ -42,16 +42,15 @@
 > **Data leaves your machine by default.** When a cloud vision model is used, Vision Router sends the image (or a derived crop), the vision prompt and related request metadata to that provider. A fresh install has an anonymous OVHcloud vision fallback enabled, so "free" and "no key" do **not** mean offline. Do not use the cloud chain for confidential, regulated or classified material. See [Data flow and strict local-only use](#data-flow-and-strict-local-only-use).
 
 > [!WARNING]
-> 📌 **Announcement (v3.0.2)**
+> 📌 **Announcement (v3.0.3)**
 >
-> **v3.0.2: hardening from a full-repository adversarial review.** Instant local describe now
-> follows the same `proxy` / `proxyHosts` decision as every other provider leg; credential
-> redaction is one shared shape list for the logs and the failure diagnostics; adapter ownership
-> no longer accepts a primitive marker; proxy host admission normalizes IPv6 spellings and makes
-> `*.host` entries work; the background stop cache can neither go stale nor vanish; and failure
-> classification, numeric settings and the retained-image walker are hardened against inherited
-> keys, hostile accessors and pathological nesting. Supported Hosts are unchanged: DSH 0.1.x from
-> `0.1.5-rc.1`, DSH 0.2.x from verified `0.2.0-rc.2`. [What’s new →](docs/releases/v3.0.2.md)
+> **v3.0.3: presentation and proxy-boundary corrections.** Restored history and empty
+> `vision_present` results no longer steal the reader’s scroll position, while fresh presented
+> images still focus correctly on both the newer phase-based Host contract and the supported DSH
+> 0.1.5 lifecycle. Cookie-domain-style `.example.com` entries now follow the same proxy-host
+> policy as `*.example.com`, and the exact DSH 0.1.7 browser smoke again follows both
+> presentation-boundary owners. CI action pins were also refreshed; runtime dependencies and
+> supported Hosts are unchanged. [What’s new →](docs/releases/v3.0.3.md)
 
 <p align="center">
   <img src="assets/vision-demo.gif" width="640" alt="Demo: paste an image, the agent locates the send button with vision_ground / vision_crop / vision_pixel_diff and answers with coordinates" />

package/README.zh.md

@@ -42,14 +42,13 @@
 > **默认配置会让数据出网。** 使用云端视觉模型时，Vision Router 会将图片（或由它裁剪出的局部）、识图提示词及相关请求元数据发送给对应供应商。新安装默认开启 OVHcloud 匿名视觉兜底，因此“免费”和“免 Key”不等于离线。涉密、涉案、内部或其他受限材料请勿使用云端链路。详见[数据流向与严格纯本地配置](#数据流向与严格纯本地配置)。
 
 > [!WARNING]
-> 📌 **公告（v3.0.2）**
+> 📌 **公告（v3.0.3）**
 >
-> **v3.0.2：一轮全仓对抗审查带来的加固。** instant 本地描述现在与其他 provider 腿遵守同一套
-> `proxy` / `proxyHosts` 判定；日志与失败诊断的凭据脱敏合并为一份共享清单；适配器归属不再接受
-> 基本类型标记；代理主机准入归一了 IPv6 写法并让 `*.host` 条目真正生效；后台停记录缓存既不残留
-> 也不会消失；失败分类、数值设置与 retained-image 遍历也分别对继承键、恶意取值器和畸形嵌套做了
-> 加固。支持的 Host 不变：DSH 0.1.x 自 `0.1.5-rc.1` 起，DSH 0.2.x 自已验证的 `0.2.0-rc.2` 起。
-> [查看完整更新 →](docs/releases/v3.0.2.md)
+> **v3.0.3：修正图片呈现与代理边界。** 恢复历史会话和空的 `vision_present` 结果不再抢走滚动位置，
+> 新生成的图片仍会在新版 phase 生命周期与受支持的 DSH 0.1.5 旧生命周期上正确聚焦；
+> cookie-domain 风格的 `.example.com` 现在与 `*.example.com` 遵守同一套代理主机规则；
+> DSH 0.1.7 精确浏览器 smoke 也重新跟随两个 presentation-boundary owner。CI Action 固定版本同步更新，
+> 运行时依赖和 Host 支持范围均不变。[查看完整更新 →](docs/releases/v3.0.3.md)
 
 <p align="center">
   <img src="assets/vision-demo.gif" width="640" alt="演示：粘贴图片，Agent 用 vision_ground / vision_crop / vision_pixel_diff 定位发送按钮并给出坐标" />

package/docs/releases/v3.0.3.md

@@ -0,0 +1,45 @@
+# v3.0.3
+
+DSH Vision Router 3.0.3 is a corrective patch on the 3.0 line. It closes the three residuals
+reported after 3.0.2 and carries two CI-action maintenance updates. There is no runtime dependency,
+settings migration, Session-format change, public-API change or Host-support-window change.
+
+## Highlights
+
+- **Presented-image focus only reacts to fresh image results (#666).** Restoring a long Session no
+  longer jumps to an old image below the viewport, and an empty `vision_present` result cannot
+  steer the reader to another card. The helper receives the exact card node instead of querying
+  the document for the last presented card.
+- **The focus rule works across the supported Host window.** DSH 0.1.7+ exposes an explicit
+  `phase`; the supported DSH 0.1.5 train does not. The client normalizes both contracts by using
+  `phase` when present and falling back to the settled `block.kind === 'tool-result'`
+  discriminant otherwise. Restored results initialize as already settled, while a live call still
+  produces the non-result → result edge.
+- **Leading-dot proxy hosts are no longer inert (#665).** `.example.com` now canonicalizes to the
+  same apex-and-subdomain policy as `*.example.com`. Dot-boundary matching remains conservative,
+  so lookalikes such as `evilexample.com` stay outside.
+- **The exact DSH 0.1.7 browser smoke follows both presentation owners again (#664).** Changes to
+  either presentation-boundary carrier now trigger the exact-Host browser regression layer.
+- **CI action maintenance (#662, #663).** `actions/download-artifact` is pinned to v8.0.1, and the
+  HOL scanner Action is pinned to v1.2.763 (plugin-scanner 3.24.2). The scanner version annotation
+  in the workflow is corrected to match the actual immutable SHA.
+
+## Compatibility
+
+Unchanged from 3.0.2:
+
+- DSH 0.1.x: `>=0.1.5-rc.1 <0.2.0-0`
+- DSH 0.2.x: `>=0.2.0-rc.2 <0.3.0-0`
+- Node.js: `^22.19.0 || >=24.0.0`
+
+## Validation
+
+The fixes carry direct regression coverage for restored/empty presentation results, the legacy
+no-`phase` Host lifecycle, leading-dot proxy admission/non-admission, and the DSH 0.1.7 workflow
+ownership paths. The release workflow re-runs the full test, build, reproducibility, type, packed
+API and Host-support gates on the exact release SHA before it creates the immutable tag or
+publishes to npm.
+
+## Upgrade
+
+Drop-in replacement for 3.0.2. No configuration or data migration is required.

package/lib/client-presentation-boundary-main.js

@@ -996,11 +996,8 @@ export const CLIENT_PRESENTATION_PRELUDE = String.raw`(function(){
     return function(){ tag.remove(); };
   }
 
-  // A presented image that mounts below the reader is the same as not being shown: the card
-  // renders next to its tool call, and a reader who is at the end of a long conversation has
-  // that call thousands of pixels above them. The card calls this on mount; it only moves the
-  // reader for content that arrived below the viewport — a card that is already visible is left
-  // alone, and so is history above the reader when a session is opened.
+  // A fresh presented image below the reader is the same as not being shown. Centre only that
+  // card; visible cards and history on either side of a restored viewport stay untouched.
   function focusPresentedCard(node, viewportHeight) {
     if (!node || typeof node.getBoundingClientRect !== 'function') return false;
     var rect = node.getBoundingClientRect();
@@ -1022,12 +1019,25 @@ export const CLIENT_PRESENTATION_PRELUDE = String.raw`(function(){
     return false;
   }
 
+  // New Hosts expose phase; supported DSH 0.1.5 instead marks settled blocks as
+  // 'tool-result'. Normalize the [phase, block kind] tuple without raising the Host floor.
+  function isPresentedResultState(state) {
+    var phase = state && state[0];
+    return phase !== undefined ? phase === 'result' : !!state && state[1] === 'tool-result';
+  }
+
+  // Restored history starts settled on both sides of the comparison. Only a live transition into
+  // an image-bearing result is fresh presentation.
+  function shouldFocusPresentedCard(previousState, state, imageCount) {
+    return imageCount > 0 &&
+      isPresentedResultState(state) &&
+      !isPresentedResultState(previousState);
+  }
+
   try {
-    globalThis.__dvrPresentFocus = function () {
-      if (typeof document === 'undefined') return false;
-      var cards = document.querySelectorAll('[data-dvr-present]');
-      var last = cards[cards.length - 1];
-      return last ? focusPresentedCard(last, globalThis.innerHeight || 0) : false;
+    globalThis.__dvrPresentFocus = function (node, previousState, state, imageCount) {
+      return shouldFocusPresentedCard(previousState, state, imageCount) &&
+        focusPresentedCard(node, globalThis.innerHeight || 0);
     };
   } catch (_) {}
 

package/lib/client.js

@@ -4756,12 +4756,11 @@ openLogFolder,
           }
         }, ownerToken)
       }
-      // Follow the app language: register our dictionaries and re-read them
-      // whenever the user switches the locale in Settings → General.
+      // Keep card copy in sync with the app locale.
       ctx.effect(() => ctx.locale.register(NS, { zh, en }), 'vision-router: card locale')
       const t = ctx.locale.bind(NS)
       const VisionPresentCard = (props) => {
-        const { block, sessionId } = props
+        const { block, sessionId, phase } = props
         const content = block && Array.isArray(block.content) ? block.content : []
         const images = content
           .filter((item) => item && item.type === 'image' && item.attachment)
@@ -4770,7 +4769,13 @@ openLogFolder,
         const imageKeySignature = JSON.stringify(imageKeys)
         const imageOwnerToken = React.useRef({}).current
         const ownedImageKeys = React.useRef(new Set())
-        React.useEffect(() => { if (globalThis.__dvrPresentFocus) globalThis.__dvrPresentFocus() }, [imageKeySignature])
+        const presentNode = React.useRef()
+        const focusState = React.useRef([phase, block?.kind])
+        React.useEffect(() => {
+          const next = [phase, block?.kind]
+          globalThis.__dvrPresentFocus?.(presentNode.current, focusState.current, next, images.length)
+          focusState.current = next
+        }, [phase, block?.kind, imageKeySignature])
         React.useEffect(() => {
           const current = new Set(imageKeys)
           for (const key of ownedImageKeys.current) {
@@ -4810,7 +4815,7 @@ openLogFolder,
         }
         return React.createElement(
           'div',
-          { 'data-dvr-present': 'true', style: { display: 'flex', flexDirection: 'column', alignItems: 'flex-start', gap: 6, padding: '4px 0' } },
+          { ref: presentNode, 'data-dvr-present': 'true', style: { display: 'flex', flexDirection: 'column', alignItems: 'flex-start', gap: 6, padding: '4px 0' } },
           parsed && typeof parsed.label === 'string' && parsed.label !== 'image'
             ? React.createElement(
                 'div',

package/lib/diagnostic-redaction.js

@@ -20,9 +20,15 @@ const CREDENTIAL_SHAPE_PATTERNS = [
   // (the log sink has always consumed it) while `,` separates what follows.
   [/\b(Authorization\s*[:=]\s*)(?:[A-Za-z][A-Za-z0-9._-]*\s+)?[^\s,]+/gi, `$1${REDACTED}`],
   [/\b(Proxy-Authorization\s*[:=]\s*)(?:[A-Za-z][A-Za-z0-9._-]*\s+)?[^\s,]+/gi, `$1${REDACTED}`],
-  // Cookies are `name=value; name=value` on one line and there is no safe way to
-  // guess which pair carries the session, so the whole header value goes.
-  [/\b((?:Set-)?Cookie\s*[:=]\s*)[^\r\n]+/gi, `$1${REDACTED}`],
+  // A real Cookie/Set-Cookie header owns the remainder of its line, so redact the
+  // whole value there. The same words also appear inside arbitrary diagnostics,
+  // URL query fragments and object dumps; those contexts must keep unrelated text.
+  [/^([ \t]*(?:Set-)?Cookie\s*:\s*)[^\r\n]+/gim, `$1${REDACTED}`],
+  // Inline colon form: redact one scalar or a semicolon-separated cookie-pair run,
+  // then stop at an unrelated comma/word so support diagnostics remain readable.
+  [/\b((?:Set-)?Cookie\s*:\s*)(?:"[^"\r\n]*"|'[^'\r\n]*'|[^;\s,&"']+(?:\s*;\s*[^=;\s,&"']+\s*=\s*(?:"[^"\r\n]*"|'[^'\r\n]*'|[^;\s,&"']+))*)/gi, `$1${REDACTED}`],
+  // Assignment/query form: bound the value at ordinary diagnostic separators.
+  [/\b((?:Set-)?Cookie\s*=\s*)(?:"[^"\r\n]*"|'[^'\r\n]*'|[^\s,&"']+)/gi, `$1${REDACTED}`],
   // URL userinfo (`scheme://user:secret@host`) without needing a URL parser, so
   // the log path — which never parsed URLs — covers it too.
   [/\b([a-z][a-z0-9+.-]*:\/\/)[^/\s@]+@/gi, `$1${REDACTED}@`],

package/lib/proxy-routing.js

@@ -9,6 +9,21 @@ export function visionProxyOverrideConfigured(config = {}) {
         : undefined;
     return typeof value === 'string' && value.trim() !== '';
 }
+function canonicalIpv6Literal(value) {
+    if (!value.includes(':'))
+        return undefined;
+    try {
+        const serialized = new URL(`http://[${value}]/`).hostname;
+        if (!serialized.startsWith('[') || !serialized.endsWith(']'))
+            return undefined;
+        return serialized.slice(1, -1).toLowerCase();
+    }
+    catch {
+        // Invalid or URL-unrepresentable literals (for example zone-id spellings)
+        // keep the existing conservative string semantics instead of broadening admission.
+        return undefined;
+    }
+}
 /**
  * Canonicalize a configured/request hostname for proxy admission only.
  * Persisted settings stay untouched; this removes DNS presentation differences
@@ -22,26 +37,30 @@ export function canonicalProxyHost(value) {
         host = host.slice(0, -1);
     if (host === '')
         return '';
-    // A `*.host` entry means "this host and its subdomains", which is exactly what the matcher
-    // already does for a bare hostname. `domainToASCII` rejects the star, so leaving it in place
-    // makes the entry equal no real hostname: the configured host silently proxies nothing.
+    // `*.host` and cookie-domain `.host` entries both mean "this host and its subdomains",
+    // which is exactly what the matcher already does for a bare hostname. Leaving either prefix
+    // in place makes the entry equal no real hostname: the configured host silently proxies nothing.
     if (host.startsWith('*.'))
         host = host.slice(2);
+    else if (host.startsWith('.'))
+        host = host.slice(1);
     if (host === '')
         return '';
-    // WHATWG URL.hostname keeps IPv6 literals bracketed. domainToASCII does not
-    // accept an IPv6 literal at all, so case-fold it directly. The brackets are
-    // pure spelling for the same address: stripping them is what lets a configured
-    // `::1` match a URL-derived `[::1]`. A bracketed non-literal (say
-    // `[example.com]`) keeps its spelling, so nothing about DNS is widened.
+    // WHATWG URL.hostname serializes IPv6 literals to one compressed spelling.
+    // Re-serialize configured literals through the same parser before comparing so
+    // equivalent legal spellings (expanded/compressed, case, leading zeroes and
+    // IPv4-mapped forms) cannot silently bypass a proxy override. A bracketed
+    // non-literal keeps its spelling, and invalid/URL-unrepresentable colon forms
+    // retain the previous lowercase-only fallback rather than widening admission.
     if (host.startsWith('[') && host.endsWith(']')) {
         const literal = host.slice(1, -1).trim();
         if (literal === '')
             return '';
-        return literal.includes(':') ? literal.toLowerCase() : host.toLowerCase();
+        return canonicalIpv6Literal(literal)
+            ?? (literal.includes(':') ? literal.toLowerCase() : host.toLowerCase());
     }
     if (host.includes(':'))
-        return host.toLowerCase();
+        return canonicalIpv6Literal(host) ?? host.toLowerCase();
     const ascii = domainToASCII(host);
     return (ascii || host).toLowerCase();
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-vision-router",
-  "version": "3.0.2",
+  "version": "3.0.3",
   "description": "Eyes for text-only DeepSeek Harness agents: built-in free vision chain + pixel-level tools. Remote vision providers, including the default free fallback, receive image content unless strict local-only use is configured.",
   "license": "MIT",
   "repository": {