From 0f74cbd6ffc3586eaab162c715ac9e4197be8214 Mon Sep 17 00:00:00 2001
From: 134 <a134@134deMacBook-Pro.local>
Date: Tue, 23 Jun 2026 18:51:21 +0800
Subject: [PATCH] fix: infer canvas agent image counts

---
 app/(chat)/api/chat/route.ts             | 15 +++++-
 lib/ai/runtime/image-count-intent.ts     | 60 ++++++++++++++++++++++++
 tests/routes/canvas-chat-helpers.test.ts | 11 +++++
 3 files changed, 85 insertions(+), 1 deletion(-)
 create mode 100644 lib/ai/runtime/image-count-intent.ts

diff --git a/app/(chat)/api/chat/route.ts b/app/(chat)/api/chat/route.ts
index 2131655fe..d25c6bb32 100644
--- a/app/(chat)/api/chat/route.ts
+++ b/app/(chat)/api/chat/route.ts
@@ -38,6 +38,7 @@ import {
   executeDirectVideoGeneration,
   generateVideo,
 } from "@/lib/ai/tools/generate-video";
+import { inferAgentImageCountFromText } from "@/lib/ai/runtime/image-count-intent";
 import {
   getAgentRequiredUploadCount,
   normalizeAgentPresetModeConfig,
@@ -2241,7 +2242,7 @@ export async function POST(request: Request) {
       !smartAutoNeedsClarification &&
       allowAgentVideoGeneration &&
       Boolean(smartAutoExecution?.shouldApplyVideoPipeline);
-    const effectiveImageParams = shouldApplyAgentImageParams
+    const mergedEffectiveImageParams = shouldApplyAgentImageParams
       ? mergeImageParamsWithAgentRecommendation({
           incoming: requestBody.imageParams,
           recommended: smartAutoRecommendedImageParams,
@@ -2249,6 +2250,17 @@ export async function POST(request: Request) {
           hasUserCustomSettings: hasUserCustomImageParams,
         })
       : requestBody.imageParams;
+    const inferredAgentImageCount =
+      shouldApplyAgentImageParams && smartAutoDetectedIntent === "image_gen"
+        ? inferAgentImageCountFromText(requestTextContent)
+        : null;
+    const effectiveImageParams =
+      inferredAgentImageCount !== null
+        ? {
+            ...(mergedEffectiveImageParams || {}),
+            imageCount: inferredAgentImageCount,
+          }
+        : mergedEffectiveImageParams;
     const effectiveVideoParams = shouldApplyAgentVideoParams
       ? mergeVideoParamsWithAgentRecommendation({
           incoming: requestBody.videoParams,
@@ -2273,6 +2285,7 @@ export async function POST(request: Request) {
       chatApiLogger.info("[Smart Auto Mode] 🎯 Effective image params:", {
         incoming: requestBody.imageParams,
         recommended: smartAutoRecommendedImageParams,
+        inferredImageCount: inferredAgentImageCount,
         merged: effectiveImageGenerationParams,
         hasUserCustomImageParams,
       });
diff --git a/lib/ai/runtime/image-count-intent.ts b/lib/ai/runtime/image-count-intent.ts
new file mode 100644
index 000000000..2cad53e2c
--- /dev/null
+++ b/lib/ai/runtime/image-count-intent.ts
@@ -0,0 +1,60 @@
+const CHINESE_NUMBER_VALUES: Record<string, number> = {
+  一: 1,
+  两: 2,
+  二: 2,
+  三: 3,
+  四: 4,
+  五: 5,
+  六: 6,
+  七: 7,
+  八: 8,
+  九: 9,
+  十: 10,
+};
+
+const COUNT_TOKEN = "([一两二三四五六七八九十]|[1-9]\\d?)";
+const COUNT_UNIT = "(?:张|幅|版|个|组)";
+
+function normalizeCountText(text: string) {
+  return String(text || "")
+    .toLowerCase()
+    .replace(/\s+/g, "")
+    .trim();
+}
+
+function parseCountToken(token?: string | null) {
+  const value = String(token || "").trim();
+  if (!value) return null;
+  if (/^\d+$/.test(value)) {
+    const numericValue = Number(value);
+    return Number.isFinite(numericValue) ? numericValue : null;
+  }
+  return CHINESE_NUMBER_VALUES[value] || null;
+}
+
+function clampImageCount(value: number, maxCount: number) {
+  return Math.min(Math.max(Math.round(value), 1), Math.max(1, maxCount));
+}
+
+export function inferAgentImageCountFromText(text: string, maxCount = 12) {
+  const normalizedText = normalizeCountText(text);
+  if (!normalizedText) return null;
+
+  const explicitCountPatterns = [
+    new RegExp(
+      `(?:再来|再出|再做|再给我|补|多来|继续生成|继续出|继续做)${COUNT_TOKEN}${COUNT_UNIT}?`
+    ),
+    new RegExp(
+      `(?:生成|创作|创建|制作|设计|做|出|画|绘制)${COUNT_TOKEN}${COUNT_UNIT}(?:图|图片|海报|头像|插画|场景|主图|产品图|商品图|电商图|壁纸|封面|logo|图标)?`
+    ),
+    new RegExp(`(?:来|出|做)${COUNT_TOKEN}${COUNT_UNIT}`),
+  ];
+
+  for (const pattern of explicitCountPatterns) {
+    const match = normalizedText.match(pattern);
+    const count = parseCountToken(match?.[1]);
+    if (count !== null && count > 0) return clampImageCount(count, maxCount);
+  }
+
+  return null;
+}
diff --git a/tests/routes/canvas-chat-helpers.test.ts b/tests/routes/canvas-chat-helpers.test.ts
index d3f5beb7a..8a054e97f 100644
--- a/tests/routes/canvas-chat-helpers.test.ts
+++ b/tests/routes/canvas-chat-helpers.test.ts
@@ -40,6 +40,7 @@ import {
 } from "@/lib/canvas-toolbar-config";
 import { shouldBypassCanvasAgentMissingTargetClarification } from "@/lib/ai/runtime/canvas-agent-clarification";
 import { hasCanvasAgentRoutingContext } from "@/lib/ai/canvas-agent-routing";
+import { inferAgentImageCountFromText } from "@/lib/ai/runtime/image-count-intent";
 import {
   isAgentModeScopedChatId,
   isScopedEmptyHistoryRequest,
@@ -158,6 +159,16 @@ import type { CanvasAgentContext } from "@/types/canvas/agent-protocol";
 import type { ChatMessage } from "@/lib/types";
 
 test.describe("canvas chat pure helpers", () => {
+  test("infers explicit agent image counts from user text", () => {
+    expect(inferAgentImageCountFromText("再出三版")).toBe(3);
+    expect(inferAgentImageCountFromText("基于这张图再来 3 张")).toBe(3);
+    expect(inferAgentImageCountFromText("生成四张主图")).toBe(4);
+    expect(inferAgentImageCountFromText("多来两张")).toBe(2);
+    expect(inferAgentImageCountFromText("生成16:9横版海报")).toBeNull();
+    expect(inferAgentImageCountFromText("提升到2K清晰度")).toBeNull();
+    expect(inferAgentImageCountFromText("再出30张", 12)).toBe(12);
+  });
+
   test("keeps canvas message empty-history access scoped and explicit", () => {
     const canvasChatId = "11111111-1111-4111-8111-111111111111";
     const agentScopedChatId = "11111111-1111-4111-8111-a93e17a60001";