@co0ontty/wand 4.72.1 → 4.72.2-beta.g6c641f4

package/dist/build-info.json

@@ -1,6 +1,6 @@
 {
-  "commit": "78b791818f725d606571f7c36f245a85388b19a5",
-  "builtAt": "2026-09-22T01:56:32.828Z",
-  "version": "4.72.1",
-  "channel": "stable"
+  "commit": "6c641f47b4ea9aac1dd76f15d123aa9475a915fb",
+  "builtAt": "2026-09-22T23:53:15.316Z",
+  "version": "4.72.2-beta.g6c641f4",
+  "channel": "beta"
 }

package/dist/message-truncator.js

@@ -3,6 +3,7 @@
  * Cards that are collapsed by default have their large results replaced
  * with a summary, and clients fetch full content on-demand via API.
  */
+import { contentHasStructuredImage } from "./structured-content.js";
 const TRUNCATION_THRESHOLD = 200;
 const SUMMARY_LENGTH = 100;
 /**
@@ -162,6 +163,11 @@ export function truncateMessagesForTransport(messages, cardDefaults, streamingTu
             const toolName = toolNameMap.get(result.tool_use_id) || "";
             if (!isToolDefaultCollapsed(toolName, cardDefaults))
                 return block;
+            // 带图片的 tool_result 永不截断：图片 part 被 JSON.stringify 截半后客户端再也
+            // 抽不出 data URI（线上表现就是「读图不显示，只剩一串 base64 残文」）。图片要么
+            // 整块送达（客户端会强制展开渲染），要么由 _truncated + 懒加载单独取回。
+            if (contentHasStructuredImage(result.content))
+                return block;
             const contentStr = getContentString(result.content);
             if (contentStr.length <= TRUNCATION_THRESHOLD)
                 return block;

package/dist/server-session-routes.js

@@ -17,6 +17,7 @@ import { sendRouteError } from "./server-request.js";
 import { registerStructuredResumeRoutes } from "./server-resume-routes.js";
 import { addHiddenSessionIds, removeHiddenSessionIds } from "./session-registry.js";
 import { enrichStructuredMessages, WAND_PROTOCOL_VERSION } from "./structured-client-protocol.js";
+import { asRecord, isStructuredImagePart } from "./structured-content.js";
 import { syncWorkspaceTaskToBoard } from "./wand-task-sync.js";
 import { buildDirectoryTree, normalizeSessionDirectory, } from "./session-directory-tree.js";
 import { projectCwdForSession, renameWorkspaceDirectory, resolveWorkspaceIdForNewSession, } from "./workspace-binding.js";
@@ -351,7 +352,7 @@ function isMergeActionAllowed(snapshot) {
 }
 export function registerSessionRoutes(app, processes, structured, storage, defaultMode, config, sessions, onSessionCreated) {
     const sessionResponseDTO = (snapshot) => {
-        const windowed = windowMessagesForTransport(enrichStructuredMessages(snapshot.messages ?? []), config.cardDefaults ?? {});
+        const windowed = windowMessagesForTransport(enrichStructuredMessages(snapshot.messages ?? [], snapshot.id), config.cardDefaults ?? {});
         return toSessionDetailDTO(snapshot, {
             messages: windowed.messages,
             messageOffset: windowed.messageOffset,
@@ -688,6 +689,41 @@ export function registerSessionRoutes(app, processes, structured, storage, defau
         }
         res.status(404).json({ error: "未找到该工具结果。" });
     });
+    // ── Inline tool image endpoint ──
+    // tool_result 里内联的 base64 图片在传输层被改写为这个 URL（避免把整张图的 base64
+    // 塞进每次快照 / WS 帧）。各端用同一套带鉴权的取图通道按需加载。
+    app.get("/api/sessions/:id/tool-images/:toolUseId/:index", (req, res) => {
+        const snapshot = sessions.get(req.params.id);
+        if (!snapshot) {
+            res.status(404).json({ error: "未找到该会话。" });
+            return;
+        }
+        const toolUseId = req.params.toolUseId;
+        const index = Number.parseInt(req.params.index, 10);
+        if (!Number.isFinite(index) || index < 0) {
+            res.status(400).json({ error: "无效的图片序号。" });
+            return;
+        }
+        for (const turn of snapshot.messages ?? []) {
+            for (const block of turn.content) {
+                if (block.type !== "tool_result" || block.tool_use_id !== toolUseId || !Array.isArray(block.content))
+                    continue;
+                const part = block.content.filter(isStructuredImagePart)[index];
+                const source = asRecord(part?.source);
+                const data = typeof source?.data === "string" ? source.data : undefined;
+                if (source?.type !== "base64" || !data) {
+                    res.status(404).json({ error: "未找到该图片。" });
+                    return;
+                }
+                res.setHeader("Content-Type", typeof source.media_type === "string" ? source.media_type : "image/png");
+                res.setHeader("Cache-Control", "private, max-age=3600");
+                res.setHeader("X-Content-Type-Options", "nosniff");
+                res.send(Buffer.from(data, "base64"));
+                return;
+            }
+        }
+        res.status(404).json({ error: "未找到该工具结果。" });
+    });
     app.post("/api/sessions/:id/worktree/merge/check", asyncRoute(async (req, res) => {
         try {
             const current = requireWorktreeSession(sessions.getLatest(req.params.id));
@@ -1016,7 +1052,7 @@ export function registerSessionRoutes(app, processes, structured, storage, defau
             const rawBudget = req.query.blockBudget;
             if (typeof rawBudget === "string" && /^\d+$/.test(rawBudget) && Number(rawBudget) > 0) {
                 const blockBudget = parseBoundedInteger(rawBudget, 1, 1, 2_000);
-                const windowed = blockWindowMessagesForTransport(enrichStructuredMessages(snapshot.messages ?? []), config.cardDefaults ?? {}, blockBudget);
+                const windowed = blockWindowMessagesForTransport(enrichStructuredMessages(snapshot.messages ?? [], snapshot.id), config.cardDefaults ?? {}, blockBudget);
                 res.json(toSessionDetailDTO(snapshot, {
                     output: transcriptOutput,
                     messages: windowed.messages,
@@ -1028,7 +1064,7 @@ export function registerSessionRoutes(app, processes, structured, storage, defau
                 return;
             }
             // 与 WS init 对齐：只回最近一窗 turn + offset/total，更早的走 /messages 翻页。
-            const windowed = windowMessagesForTransport(enrichStructuredMessages(snapshot.messages ?? []), config.cardDefaults ?? {});
+            const windowed = windowMessagesForTransport(enrichStructuredMessages(snapshot.messages ?? [], snapshot.id), config.cardDefaults ?? {});
             res.json(toSessionDetailDTO(snapshot, {
                 output: transcriptOutput,
                 messages: windowed.messages,
@@ -1048,7 +1084,7 @@ export function registerSessionRoutes(app, processes, structured, storage, defau
             res.status(404).json({ error: "未找到该会话，可能已被删除。" });
             return;
         }
-        const all = enrichStructuredMessages(snapshot.messages ?? []);
+        const all = enrichStructuredMessages(snapshot.messages ?? [], snapshot.id);
         const total = all.length;
         // 块级翻页（iOS）：?turn=<i>&blockOffset=<当前 leading 偏移>&blockLimit=<N>
         // 取该 turn 的 [start, blockOffset) 段（start = max(0, blockOffset - blockLimit)）。
@@ -1069,7 +1105,7 @@ export function registerSessionRoutes(app, processes, structured, storage, defau
             const blocks = enrichStructuredMessages([{
                     ...turn,
                     content: sliceTurnBlocksForTransport(turn, blockStart, blockEnd, config.cardDefaults ?? {}),
-                }])[0].content;
+                }], snapshot.id)[0].content;
             res.json({ wandProtocolVersion: WAND_PROTOCOL_VERSION, turnIndex, blocks, blockOffset: blockStart, blockTotal });
             return;
         }
@@ -1078,7 +1114,7 @@ export function registerSessionRoutes(app, processes, structured, storage, defau
         const limit = Math.min(Math.max(Number.isFinite(rawLimit) ? rawLimit : 40, 1), 200);
         const offset = Math.min(Math.max(Number.isFinite(rawOffset) ? rawOffset : 0, 0), total);
         const end = Math.min(offset + limit, total);
-        const slice = enrichStructuredMessages(truncateMessagesForTransport(all.slice(offset, end), config.cardDefaults ?? {}));
+        const slice = enrichStructuredMessages(truncateMessagesForTransport(all.slice(offset, end), config.cardDefaults ?? {}), snapshot.id);
         res.json({ wandProtocolVersion: WAND_PROTOCOL_VERSION, messages: slice, offset, total });
     });
     app.post("/api/sessions/:id/resume", asyncRoute(async (req, res) => {

package/dist/session-transport.js

@@ -87,7 +87,7 @@ export function toSessionDetailDTO(snapshot, options = {}) {
         outputOffset,
         outputTotal: rawOutput.length,
         outputTruncated: outputOffset > 0,
-        ...(options.messages !== undefined ? { messages: enrichStructuredMessages(options.messages) } : {}),
+        ...(options.messages !== undefined ? { messages: enrichStructuredMessages(options.messages, snapshot.id) } : {}),
         ...(options.messageOffset !== undefined ? { messageOffset: options.messageOffset } : {}),
         ...(options.messageTotal !== undefined ? { messageTotal: options.messageTotal } : {}),
         ...(options.leadingBlockOffset !== undefined ? { leadingBlockOffset: options.leadingBlockOffset } : {}),

package/dist/structured-client-protocol.d.ts

@@ -3,5 +3,7 @@ export declare const WAND_PROTOCOL_VERSION = 2;
 /**
  * Add Wand-owned semantics without mutating persisted provider blocks.
  * This is the external interface consumed by every client.
+ *
+ * 传 sessionId 时额外把内联图片改写成取图 URL（见 rewriteInlineToolImageUrls）。
  */
-export declare function enrichStructuredMessages(messages: ConversationTurn[]): ConversationTurn[];
+export declare function enrichStructuredMessages(messages: ConversationTurn[], sessionId?: string): ConversationTurn[];

package/dist/structured-client-protocol.js

@@ -1,4 +1,4 @@
-import { asRecord } from "./structured-content.js";
+import { asRecord, isStructuredImagePart } from "./structured-content.js";
 export const WAND_PROTOCOL_VERSION = 2;
 function text(value) {
     return typeof value === "string" && value ? value : undefined;
@@ -190,11 +190,54 @@ function tasksFromSegment(messages, start, end) {
         targetId,
     };
 }
+/**
+ * 把 tool_result 里的内联 base64 图片换成「按会话取图」的 URL。
+ *
+ * 为什么不能把 base64 直接下发：读一张截图动辄 1–3MB，base64 后更大，会撑爆移动端
+ * 单帧上限（iOS 默认 1MiB）导致反复断连；而且同一张图在每条快照里重复传。换成 URL
+ * 后各端用同一套带鉴权的取图通道按需加载（Web 走 <img>，原生走各自 ImageLoader）。
+ * source 本来就是 url 的图片不动，保持幂等。
+ */
+function rewriteInlineToolImageUrls(messages, sessionId) {
+    return messages.map((turn) => {
+        let turnChanged = false;
+        const content = turn.content.map((block) => {
+            if (block.type !== "tool_result" || !Array.isArray(block.content))
+                return block;
+            let imageIndex = -1;
+            let blockChanged = false;
+            const parts = block.content.map((part) => {
+                if (!isStructuredImagePart(part))
+                    return part;
+                imageIndex += 1;
+                const source = asRecord(part.source);
+                if (source?.type !== "base64" || typeof source.data !== "string")
+                    return part;
+                blockChanged = true;
+                return {
+                    type: "image",
+                    source: {
+                        type: "url",
+                        url: `/api/sessions/${encodeURIComponent(sessionId)}/tool-images/${encodeURIComponent(block.tool_use_id)}/${imageIndex}`,
+                        ...(typeof source.media_type === "string" ? { media_type: source.media_type } : {}),
+                    },
+                };
+            });
+            if (!blockChanged)
+                return block;
+            turnChanged = true;
+            return { ...block, content: parts };
+        });
+        return turnChanged ? { ...turn, content } : turn;
+    });
+}
 /**
  * Add Wand-owned semantics without mutating persisted provider blocks.
  * This is the external interface consumed by every client.
+ *
+ * 传 sessionId 时额外把内联图片改写成取图 URL（见 rewriteInlineToolImageUrls）。
  */
-export function enrichStructuredMessages(messages) {
+export function enrichStructuredMessages(messages, sessionId) {
     const enriched = stampDerivedSubagents(messages).map((turn) => ({
         ...turn,
         content: turn.content.map((block) => {
@@ -231,5 +274,5 @@ export function enrichStructuredMessages(messages) {
         }
         segmentStart = i;
     }
-    return enriched;
+    return sessionId ? rewriteInlineToolImageUrls(enriched, sessionId) : enriched;
 }

package/dist/structured-content.d.ts

@@ -1,7 +1,24 @@
 /** 把 unknown 收敛成普通对象；数组 / null / 原始值都返回 null。 */
 export declare function asRecord(value: unknown): Record<string, unknown> | null;
-/** Preserve both Responses content parts and arbitrary structured tool output. */
-export declare function normalizeStructuredToolResultContent(content: unknown): string | Array<{
+export type StructuredContentPart = {
     type: string;
     [key: string]: unknown;
-}>;
+};
+/**
+ * 各家 provider 的图片 part 形态不同：
+ *   · Anthropic：`{ type:"image", source:{ type:"base64", media_type, data } }` / `source:{ type:"url", url }`
+ *   · OpenAI Responses：`{ type:"image_url", image_url:{ url } }` / `image_url:"…"`
+ *   · Pi：`{ type:"image", data, mimeType }`
+ * 只要带图片语义就算，具体归一化交给 canonicalizeStructuredImagePart。
+ */
+export declare function isStructuredImagePart(value: unknown): boolean;
+/**
+ * 统一成客户端认识的规范形态（Web / Android / iOS / macOS 都按
+ * `{ type:"image", source:{ type:"base64"|"url", … } }` 解析）。识别不了返回 null，
+ * 由调用方原样保留，避免丢信息。
+ */
+export declare function canonicalizeStructuredImagePart(value: unknown): StructuredContentPart | null;
+/** content（字符串 / part 数组）里是否含图片 part。 */
+export declare function contentHasStructuredImage(content: unknown): boolean;
+/** Preserve both Responses content parts and arbitrary structured tool output. */
+export declare function normalizeStructuredToolResultContent(content: unknown): string | StructuredContentPart[];

package/dist/structured-content.js

@@ -4,12 +4,91 @@ export function asRecord(value) {
         ? value
         : null;
 }
+function stringValue(value) {
+    return typeof value === "string" && value.length > 0 ? value : undefined;
+}
+/**
+ * 各家 provider 的图片 part 形态不同：
+ *   · Anthropic：`{ type:"image", source:{ type:"base64", media_type, data } }` / `source:{ type:"url", url }`
+ *   · OpenAI Responses：`{ type:"image_url", image_url:{ url } }` / `image_url:"…"`
+ *   · Pi：`{ type:"image", data, mimeType }`
+ * 只要带图片语义就算，具体归一化交给 canonicalizeStructuredImagePart。
+ */
+export function isStructuredImagePart(value) {
+    const part = asRecord(value);
+    if (!part)
+        return false;
+    if (part.type === "image_url")
+        return !!imageUrlOf(part);
+    if (part.type !== "image")
+        return false;
+    const source = asRecord(part.source);
+    if (source && (stringValue(source.data) || stringValue(source.url)))
+        return true;
+    return !!(stringValue(part.data) || stringValue(part.url));
+}
+function imageUrlOf(part) {
+    const raw = part.image_url;
+    if (typeof raw === "string")
+        return stringValue(raw);
+    return stringValue(asRecord(raw)?.url);
+}
+/**
+ * 统一成客户端认识的规范形态（Web / Android / iOS / macOS 都按
+ * `{ type:"image", source:{ type:"base64"|"url", … } }` 解析）。识别不了返回 null，
+ * 由调用方原样保留，避免丢信息。
+ */
+export function canonicalizeStructuredImagePart(value) {
+    const part = asRecord(value);
+    if (!part)
+        return null;
+    if (part.type === "image_url") {
+        const url = imageUrlOf(part);
+        return url ? { type: "image", source: { type: "url", url } } : null;
+    }
+    if (part.type !== "image")
+        return null;
+    const source = asRecord(part.source);
+    if (source) {
+        const data = stringValue(source.data);
+        if (source.type === "base64" && data) {
+            return {
+                type: "image",
+                source: { type: "base64", media_type: stringValue(source.media_type) ?? "image/png", data },
+            };
+        }
+        const url = stringValue(source.url);
+        if (url)
+            return { type: "image", source: { type: "url", url } };
+    }
+    const data = stringValue(part.data);
+    if (data) {
+        const mediaType = stringValue(part.mimeType)
+            ?? stringValue(part.mime_type)
+            ?? stringValue(source?.media_type)
+            ?? "image/png";
+        return { type: "image", source: { type: "base64", media_type: mediaType, data } };
+    }
+    const url = stringValue(part.url);
+    if (url)
+        return { type: "image", source: { type: "url", url } };
+    return null;
+}
+/** content（字符串 / part 数组）里是否含图片 part。 */
+export function contentHasStructuredImage(content) {
+    if (!Array.isArray(content))
+        return false;
+    return content.some((part) => isStructuredImagePart(part));
+}
 /** Preserve both Responses content parts and arbitrary structured tool output. */
 export function normalizeStructuredToolResultContent(content) {
     if (typeof content === "string")
         return content;
     if (Array.isArray(content)) {
-        const parts = content.filter((item) => !!item && typeof item === "object" && typeof item.type === "string");
+        const parts = content
+            .filter((item) => !!item && typeof item === "object" && typeof item.type === "string")
+            // 图片 part 归一化后客户端才能跨端渲染；非图片 part 原样保留。
+            .map((part) => canonicalizeStructuredImagePart(part) ?? part);
         if (parts.length === content.length)
             return parts;
         try {

package/dist/structured-pi-adapter.js

@@ -1,6 +1,6 @@
 import { spawn } from "node:child_process";
 import { startStructuredCli } from "./structured-exec-pump.js";
-import { asRecord } from "./structured-content.js";
+import { asRecord, normalizeStructuredToolResultContent } from "./structured-content.js";
 import { thinkingEffortToPiLevel } from "./structured-provider-common.js";
 export function isMissingPiSession(error, sessionId) {
     return !!sessionId && !!error?.includes(`No session found matching '${sessionId}'`);
@@ -17,6 +17,22 @@ function textContent(value) {
         return item.text;
     return textContent(item.content);
 }
+/**
+ * Pi 的工具结果是 `{ content: [ text…, image… ] }`。图片 part 不能像以前那样只抽文本
+ * 丢掉——读图 / 截图全靠它，统一交给 normalizeStructuredToolResultContent 归一化。
+ * 纯文本结果仍压成字符串，保持既有形态。
+ */
+function piToolResultContent(value) {
+    const record = asRecord(value);
+    const raw = record && "content" in record ? record.content : value;
+    const normalized = normalizeStructuredToolResultContent(raw);
+    if (typeof normalized === "string")
+        return normalized;
+    if (normalized.length > 0 && normalized.every((part) => part.type === "text" && typeof part.text === "string")) {
+        return normalized.map((part) => part.text).filter(Boolean).join("\n");
+    }
+    return normalized;
+}
 export function buildPiArgs(session, prompt) {
     const args = ["--mode", "json", "--print"];
     const model = session.selectedModel?.trim();
@@ -141,7 +157,7 @@ export function applyPiEvent(state, event) {
     }
     if (event.type === "tool_execution_end") {
         const id = typeof event.toolCallId === "string" ? event.toolCallId : "unknown";
-        state.blocks.push({ type: "tool_result", tool_use_id: id, content: textContent(event.result), is_error: event.isError === true });
+        state.blocks.push({ type: "tool_result", tool_use_id: id, content: piToolResultContent(event.result), is_error: event.isError === true });
     }
     if (event.type === "message_end" || event.type === "turn_end") {
         const message = asRecord(event.message);

package/dist/structured-session-manager.js

@@ -207,7 +207,7 @@ function buildStructuredOutputPayload(snapshot) {
     return {
         wandProtocolVersion: WAND_PROTOCOL_VERSION,
         output: snapshot.output,
-        messages: snapshot.messages ? enrichStructuredMessages(snapshot.messages) : undefined,
+        messages: snapshot.messages ? enrichStructuredMessages(snapshot.messages, snapshot.id) : undefined,
         queuedMessages: snapshot.queuedMessages,
         queuedMessageSkills: snapshot.queuedMessageSkills,
         sessionKind: "structured",
@@ -259,7 +259,7 @@ function buildIncrementalStructuredPayload(snapshot, cardDefaults) {
     const messages = snapshot.messages ?? [];
     // Derive semantics from the complete current turn before taking the
     // incremental tail; TaskCreate results can live in an earlier provider frame.
-    const clientMessages = enrichStructuredMessages(messages);
+    const clientMessages = enrichStructuredMessages(messages, snapshot.id);
     const lastTurn = clientMessages.length > 0 ? clientMessages[clientMessages.length - 1] : undefined;
     // Streaming turn (index 0 here) is preserved verbatim; truncation only kicks
     // in if the live response is already bigger than the transport threshold,