dsh-codex-connect 0.1.0-alpha.4.44 → 0.1.0-alpha.4.45

package/README.i18n.yaml

@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   git hash-object README.md docs/README.zh.md
-README.md: 38b900237ec1219376edba97fb1265903d48b9b7
-docs/README.zh.md: 9ae847d4b82d0af3eaf63819119dd1f394a075a4
+README.md: 1944b620eca818c50a7fa94e694706f0cd43f654
+docs/README.zh.md: dec04b07db9a90fd2babd153073716d9d9b69409

package/README.md

@@ -86,7 +86,7 @@ All options below are off on a fresh installation. Edit them in **Settings → P
 
 With `enableReserveFallback: true`, the account UI and agent routing share one identity-bound quota state. Background refresh follows the returned quota windows while recently in use; fresh state is reused across agent steps. Ordinary quota reads also share the cache, even with Reserve disabled. The plugin enters `gpt-reserve` only with complete, non-FedRAMP account/user identity and backend Luna Reserve authorization, then restores the session's previous model and reasoning effort after confirmed ordinary-usage recovery. Reserve has its own allowance, is hidden from the model picker, and is not unlimited. The backend decides eligibility; reset times alone do not authorize a switch. This version supports known `gpt-5.6-luna` metadata only. See [Luna Reserve fallback](docs/reference.md#luna-reserve-fallback) for refresh, identity, and verification limits.
 
-Use the image generation capability included with your current GPT subscription. Generated originals are stored separately from attachment previews; disabling the capability or uninstalling the plugin does not delete them. See [Configuration and recovery](docs/reference.md#search-and-image-tools) for storage and access rules.
+Use the image generation capability included with your current GPT subscription. Successful images appear below the conversation answer without opening its processing details; the original tool card remains available there. Generated originals are stored separately from attachment previews; disabling the capability or uninstalling the plugin does not delete them. See [Configuration and recovery](docs/reference.md#search-and-image-tools) for storage and access rules.
 
 Auto-review operates after Harness policy requires approval; it does not bypass that policy. See [Auto-review behavior](docs/auto-review.md) before enabling it.
 

package/docs/README.zh.md

@@ -86,7 +86,7 @@ Alpha 4.40 还会在旧提供方目录缺失时补充 `gpt-6-sol` 和 `gpt-6-lun
 
 启用 `enableReserveFallback: true` 后，账户 UI 和 agent 路由共用一份绑定身份的额度状态，最近有使用需求时按服务端返回的额度窗口后台刷新；有效状态可跨 agent step 复用。普通额度读取也共用缓存，即使 Reserve 已关闭。只有身份完整、非 FedRAMP 且服务端授权时，插件才进入 `gpt-reserve`；普通额度确认恢复后，切回该会话先前的模型和推理强度。Reserve 有自己的额度，不出现在模型选择器中，也不是无限额度。资格由服务端决定，重置时间本身不授权切换。当前版本只支持已知的 `gpt-5.6-luna` 元数据。刷新、身份和验证限制见 [Luna Reserve 回退](reference.zh.md#luna-reserve-回退)。
 
-使用你当前 GPT 订阅计划提供的图片生成能力。生成原文件与附件预览分开保存；关闭能力或卸载插件不会删除这些文件。存储和访问规则见[配置与恢复](reference.zh.md#搜索与图片工具)。
+使用你当前 GPT 订阅计划提供的图片生成能力。成功生成的图片会显示在对话回复下方，无需展开处理过程；原工具卡片仍保留在处理过程里。生成原文件与附件预览分开保存；关闭能力或卸载插件不会删除这些文件。存储和访问规则见[配置与恢复](reference.zh.md#搜索与图片工具)。
 
 自动审查在 Harness 策略判定需要审批后执行，不会绕过该策略。启用前请阅读[自动审查行为](auto-review.zh.md)。
 

package/docs/agent-notes/image-results-visible-answer.md

@@ -0,0 +1,7 @@
+# Image results in the answer
+
+DSH 0.1.7-rc.1 groups Tool calls inside a completed Turn's collapsed processing disclosure. Codex Connect rendered generated images only through its keyed Tool view, so a successful image could remain invisible until the reader expanded that disclosure. The final Assistant message contains text rather than a second image block; the image bytes and preview reference were not missing.
+
+The browser plugin now adds a `conversation.chat.turnTail` entry. That Host-owned location is independent of the processing disclosure. It observes only the current Turn's Tool rows, including nested PTC results and result-only history windows, and reuses the same validated image presentation and complete result card. The original Tool view remains in processing details. The plugin does not write a duplicate Assistant message or copy image bytes into the Session log.
+
+The regression test first failed because the completed Turn had no visible gallery. It now checks direct and nested successful results, filtering of failed and unrelated calls, closing-sequence exclusion, and the visible gallery with no expanded processing state. A headless Chromium check confirms the image and download action are visible outside any hidden processing content. Acceptance on an isolated live DSH instance remains separate from these synthetic tests.

package/docs/design.md

@@ -40,6 +40,8 @@ When `enableImageTool: true`, `view_image` is registered only after tools, files
 
 Each remote-image redirect hop has one 30-second deadline covering DNS, connection, and body consumption. Cancellation settles the caller's wait immediately; an operating-system DNS lookup may still finish later, but its result cannot start a new HTTP request after cancellation or timeout.
 
+Successful Codex image-generation Tool results also render in the independent completed-Turn tail. DSH groups Tool calls into a collapsed processing disclosure, so the keyed Tool card alone cannot make generated images visible with the answer. The tail reads only this Turn's settled Tool rows, including nested PTC calls and restored history, and uses the same validated result metadata, session-bound preview loader, Lightbox, and download controls as the Tool card. Failed, malformed, unrelated, or post-answer results do not enter the visible image row. The Tool card remains available in processing details.
+
 ## Conflicts and diagnostics
 
 Before registration the plugin checks current provider ids. An existing `openai-codex` adapter produces a focused message naming the likely legacy-bundle or manual-provider cause. The boot-free CLI `doctor` reports package/runtime version, OAuth path metadata, capability defaults, and safe conflict guidance without returning auth content.

package/docs/release-notes/alpha-4.45.md

@@ -0,0 +1,21 @@
+# Codex Connect Alpha 4.45 — generated images visible in answers
+
+## English
+
+Generated images now appear below a completed answer without opening the collapsed processing details. The existing Tool result remains available inside those details, including its original-image download and follow-up controls. The answer view uses the same validated image result and attachment access; it does not generate a second image, alter the provider request, or duplicate image bytes in the Session log.
+
+This release targets DeepSeek Harness `0.1.7-rc.1`. The fix passed local unit and Chromium tests, and the reporter confirmed the visible result on an isolated 3081 instance. That review did not send a new image-generation request or validate other DSH versions. Preparing this package does not upgrade existing installations; publication and the npm `latest` promotion are separately verified steps.
+
+After npm publication is confirmed, install the exact version with:
+
+```sh
+dsh plugin --profile web add dsh-codex-connect@0.1.0-alpha.4.45
+```
+
+## 中文
+
+生成的图片现在会显示在已完成回复的下方，不必展开折叠的“处理过程”。原有 Tool 结果仍保留在处理详情中，包括原图下载和后续操作。回复区域复用同一份已验证的图片结果与附件访问，不会再次生图、改变 Provider 请求，也不会在 Session 日志中复制图片数据。
+
+本版针对 DeepSeek Harness `0.1.7-rc.1`。修复已通过本地单元测试和 Chromium 测试，并由报告者在隔离的 3081 实例上确认图片可见。该次验收没有发起新的真实生图请求，也不代表验证了其他 DSH 版本。准备安装包不会升级现有实例；发布和 npm `latest` 提升分别核验。
+
+确认 npm 发布后，可使用上面的精确版本命令安装。

package/lib/bin.js

@@ -765,7 +765,7 @@ async function detectCompatibility(options = {}) {
 }
 //#endregion
 //#region src/version.ts
-const CODEX_CONNECT_VERSION = "0.1.0-alpha.4.44";
+const CODEX_CONNECT_VERSION = "0.1.0-alpha.4.45";
 //#endregion
 //#region src/doctor.ts
 /** Secret-free diagnostics and duplicate-provider guidance. */

package/lib/client.js

@@ -5376,6 +5376,35 @@ window.__ModuleLoader__.load({
 			};
 		}
 		//#endregion
+		//#region src/client/image-result-presentation.ts
+		const IMAGE_TOOL = "codex_connect_image_generate";
+		function promptFromArgs(raw) {
+			try {
+				const value = JSON.parse(raw);
+				if (typeof value !== "object" || value === null || Array.isArray(value)) return void 0;
+				const prompt = value.prompt;
+				if (typeof prompt !== "string") return void 0;
+				const trimmed = prompt.trim();
+				return trimmed.length > 0 && trimmed.length <= 32e3 ? trimmed : void 0;
+			} catch {
+				return;
+			}
+		}
+		/** Return the bounded prompt that belongs to this image Tool row. */
+		function promptForImageToolBlock(block) {
+			if (!("kind" in block)) return block.phase === "start" ? promptFromArgs(block.argsRaw) : void 0;
+			const decoded = decodeImagePresentationMeta(block.meta);
+			if (decoded !== void 0) return decoded.prompt;
+			return block.call === null ? void 0 : promptFromArgs(block.call.argsRaw);
+		}
+		/** Only a successful image Tool result can enter the visible answer gallery. */
+		function imagePresentationForResult(block) {
+			if (block.isError || block.call !== null && block.call.name !== IMAGE_TOOL) return void 0;
+			if (block.meta !== void 0) return decodeImagePresentationMeta(block.meta);
+			const prompt = promptForImageToolBlock(block);
+			return prompt === void 0 ? void 0 : decodeImageResultContent(block.content, prompt);
+		}
+		//#endregion
 		//#region src/client/CodexImageGallery.tsx
 		/** Self-contained result-image gallery for the tool-call view.
 		*
@@ -5887,30 +5916,6 @@ window.__ModuleLoader__.load({
 		function contentText(content) {
 			for (const block of content) if (typeof block === "object" && block !== null && block.type === "text" && typeof block.text === "string") return block.text;
 		}
-		function presentation(block) {
-			if (!("kind" in block) || block.kind !== "tool-result" || block.isError) return void 0;
-			if (block.meta !== void 0) return decodeImagePresentationMeta(block.meta);
-			const prompt = promptFor(block);
-			return prompt === void 0 ? void 0 : decodeImageResultContent(block.content, prompt);
-		}
-		function promptFromArgs(raw) {
-			try {
-				const value = JSON.parse(raw);
-				if (typeof value !== "object" || value === null || Array.isArray(value)) return void 0;
-				const prompt = value.prompt;
-				if (typeof prompt !== "string") return void 0;
-				const trimmed = prompt.trim();
-				return trimmed.length > 0 && trimmed.length <= 32e3 ? trimmed : void 0;
-			} catch {
-				return;
-			}
-		}
-		function promptFor(block) {
-			if (!("kind" in block)) return promptFromArgs(block.argsRaw);
-			const decoded = decodeImagePresentationMeta(block.meta);
-			if (decoded !== void 0) return decoded.prompt;
-			return block.call === null ? void 0 : promptFromArgs(block.call.argsRaw);
-		}
 		function useSessionActions(sessionId, sessions) {
 			const [pending, setPending] = (0, react.useState)(null);
 			const [failed, setFailed] = (0, react.useState)(false);
@@ -6246,8 +6251,8 @@ window.__ModuleLoader__.load({
 			const load = useImageLoader(sessionId, sessions);
 			const sessionActions = useSessionActions(sessionId, sessions);
 			const galleryLabels = (0, react.useMemo)(() => labels(t), [t]);
-			const prompt = promptFor(block);
-			const decoded = (0, react.useMemo)(() => presentation(block), [block]);
+			const prompt = promptForImageToolBlock(block);
+			const decoded = (0, react.useMemo)(() => "kind" in block ? imagePresentationForResult(block) : void 0, [block]);
 			if (!("kind" in block)) return /* @__PURE__ */ (0, react_jsx_runtime.jsx)(ResponsiveCard, {
 				label: t("generating"),
 				visual: /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
@@ -6339,6 +6344,32 @@ window.__ModuleLoader__.load({
 					children: t("unknownResult")
 				})]
 			});
+			return /* @__PURE__ */ (0, react_jsx_runtime.jsx)(ImageResultCard, {
+				decoded,
+				sessionId,
+				load,
+				sessionActions,
+				galleryLabels,
+				t
+			});
+		}
+		/** Render the same complete result controls in the independent answer tail. */
+		function CodexImageTurnResult({ block, sessionId, sessions, t }) {
+			const load = useImageLoader(sessionId, sessions);
+			const sessionActions = useSessionActions(sessionId, sessions);
+			const galleryLabels = (0, react.useMemo)(() => labels(t), [t]);
+			const decoded = (0, react.useMemo)(() => imagePresentationForResult(block), [block]);
+			if (decoded === void 0) return null;
+			return /* @__PURE__ */ (0, react_jsx_runtime.jsx)(ImageResultCard, {
+				decoded,
+				sessionId,
+				load,
+				sessionActions,
+				galleryLabels,
+				t
+			});
+		}
+		function ImageResultCard({ decoded, sessionId, load, sessionActions, galleryLabels, t }) {
 			return /* @__PURE__ */ (0, react_jsx_runtime.jsx)(ResponsiveCard, {
 				label: t("completed"),
 				visual: /* @__PURE__ */ (0, react_jsx_runtime.jsxs)(react_jsx_runtime.Fragment, { children: [/* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
@@ -6449,8 +6480,53 @@ window.__ModuleLoader__.load({
 			});
 		}
 		//#endregion
+		//#region src/client/CodexImageTurnTail.tsx
+		/** Visible generated-image results after a completed conversation turn. */
+		const resultStack = {
+			display: "grid",
+			gap: 12,
+			minWidth: 0
+		};
+		/** Keep successful image results from this turn, including nested PTC calls. */
+		function selectTurnImageResults(rows, closingSeq) {
+			const images = [];
+			for (const row of rows) {
+				const pending = [row.root];
+				const visited = /* @__PURE__ */ new Set();
+				while (pending.length > 0) {
+					const block = pending.pop();
+					if (block === void 0 || visited.has(block.callId)) continue;
+					visited.add(block.callId);
+					if ("kind" in block && block.seq <= closingSeq && imagePresentationForResult(block) !== void 0) images.push(block);
+					for (let index = block.subCalls.length - 1; index >= 0; index--) {
+						const child = block.subCalls[index];
+						if (child !== void 0) pending.push(child);
+					}
+				}
+			}
+			return images.sort((left, right) => left.seq - right.seq);
+		}
+		/** Put generated images in the independent answer tail, outside Turn-process disclosure. */
+		function CodexImageTurnTail({ turn, seq, sessionId, useChat, sessions, t }) {
+			const source = useChat((chat) => chat.nodes.turnDataSource(turn.turn, "tool-call"));
+			const rows = (0, react.useSyncExternalStore)(source.subscribe, source.getSnapshot, source.getSnapshot);
+			const results = (0, react.useMemo)(() => selectTurnImageResults(rows, seq), [rows, seq]);
+			if (results.length === 0) return null;
+			return /* @__PURE__ */ (0, react_jsx_runtime.jsx)("section", {
+				"data-turn-image-results": "",
+				"aria-label": t("completed"),
+				style: resultStack,
+				children: results.map((block) => /* @__PURE__ */ (0, react_jsx_runtime.jsx)(CodexImageTurnResult, {
+					block,
+					sessionId,
+					sessions,
+					t
+				}, block.callId))
+			});
+		}
+		//#endregion
 		//#region src/version.ts
-		const CODEX_CONNECT_VERSION = "0.1.0-alpha.4.44";
+		const CODEX_CONNECT_VERSION = "0.1.0-alpha.4.45";
 		//#endregion
 		//#region src/client/OpenAICodexModelsCard.tsx
 		/** Compact Models account entry with quota disclosure and shared configuration. */
@@ -7120,6 +7196,13 @@ window.__ModuleLoader__.load({
 				locale: namespace,
 				inject: () => ({ sessions: ctx.sessions })
 			}, CodexImageToolView));
+			ctx.slots.inject("conversation.chat.turnTail", () => ctx.slots.register({
+				name: "conversation.chat.turnTail",
+				id: "codex-connect-generated-images",
+				order: 20,
+				locale: namespace,
+				inject: () => ({ sessions: ctx.sessions })
+			}, CodexImageTurnTail));
 			ctx.inject(["slots", "modelDirectories"], (scope) => {
 				scope.slots.inject("conversation.input.right", () => scope.slots.register({
 					name: "conversation.input.right",

package/lib/index.js

@@ -5294,7 +5294,7 @@ function registerOpenAICodexOriginalImageRoute(ctx, trustedOrigins, assets) {
 }
 //#endregion
 //#region src/version.ts
-const CODEX_CONNECT_VERSION = "0.1.0-alpha.4.44";
+const CODEX_CONNECT_VERSION = "0.1.0-alpha.4.45";
 //#endregion
 //#region src/doctor.ts
 /** Secret-free diagnostics and duplicate-provider guidance. */

package/package.json

@@ -2,7 +2,7 @@
   "name": "dsh-codex-connect",
   "displayName": "Codex Connect",
   "description": "ChatGPT OAuth and Codex models for DeepSeek Harness.",
-  "version": "0.1.0-alpha.4.44",
+  "version": "0.1.0-alpha.4.45",
   "author": "Frank Song",
   "contributors": [
     "Yan-Zero (original dsh-codex author)"
@@ -97,6 +97,7 @@
         "@deepseek-ai/dsh-api-session-controller",
         "@deepseek-ai/dsh-client-ui-layout",
         "@deepseek-ai/dsh-client-ui-conversation",
+        "@deepseek-ai/dsh-client-ui-chat",
         "@deepseek-ai/dsh-client-ui-model-selection",
         "@deepseek-ai/dsh-client-ui-settings",
         "@deepseek-ai/dsh-client-ui-settings-plugins",
@@ -111,6 +112,7 @@
     "@deepseek-ai/dsh-agent": "0.1.7-rc.1",
     "@deepseek-ai/dsh-atomic-write": "0.1.7-rc.1",
     "@deepseek-ai/dsh-attachment": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-client-ui-chat": "0.1.7-rc.1",
     "@deepseek-ai/dsh-client-ui-layout": "0.1.7-rc.1",
     "@deepseek-ai/dsh-client-ui-primitives": "0.1.7-rc.1",
     "@deepseek-ai/dsh-client-ui-tool": "0.1.7-rc.1",
@@ -153,6 +155,7 @@
     "@deepseek-ai/dsh-client-connection": "0.1.7-rc.1",
     "@deepseek-ai/dsh-client-locale": "0.1.7-rc.1",
     "@deepseek-ai/dsh-client-store": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-client-ui-chat": "0.1.7-rc.1",
     "@deepseek-ai/dsh-client-ui-conversation": "0.1.7-rc.1",
     "@deepseek-ai/dsh-client-ui-layout": "0.1.7-rc.1",
     "@deepseek-ai/dsh-client-ui-model-selection": "0.1.7-rc.1",