From 0f3b562dc7dca58f27376466e6a8640ce49417cc Mon Sep 17 00:00:00 2001
From: Your Name <your.email@example.com>
Date: Thu, 19 Feb 2026 02:07:38 +0800
Subject: [PATCH] fix(agent): relax vision model detection for image analysis

---
 lib/ai/auto-agent-graph.ts | 45 +++++++++++++++++++----
 lib/ai/smart-analyzer.ts   | 74 ++++++++++++++++++++++++++++++--------
 2 files changed, 97 insertions(+), 22 deletions(-)

diff --git a/lib/ai/auto-agent-graph.ts b/lib/ai/auto-agent-graph.ts
index 06308904a..7eb4eec99 100644
--- a/lib/ai/auto-agent-graph.ts
+++ b/lib/ai/auto-agent-graph.ts
@@ -487,8 +487,44 @@ type EnabledModelWithProvider = Awaited<
   ReturnType<typeof import("@/lib/db/queries")["getEnabledAiModelsWithProvider"]>
 >[number];
 
+function isLikelyVisionCapableModel(item: EnabledModelWithProvider) {
+  if (!item.model.isEnabled || item.model.isSubModel) return false;
+  if (item.model.supportsVision) return true;
+  if (isAiModelMultimodal(item.model)) return true;
+
+  const providerType = item.provider.type || "";
+  const modelText = `${item.model.displayName || ""} ${item.model.modelId || ""}`.toLowerCase();
+
+  if (providerType === "volcano-multimodal") return true;
+
+  if (
+    providerType === "zhipu" &&
+    /(glm[-_.]?4v|glm[-_.]?4\.?6v|vision|多模态)/i.test(modelText)
+  ) {
+    return true;
+  }
+
+  if (
+    providerType === "alibaba" &&
+    /(qwen|vl|vision|通义|多模态)/i.test(modelText)
+  ) {
+    return true;
+  }
+
+  if (
+    ["custom", "openai", "local"].includes(providerType) &&
+    /(doubao[-_.]?seed[-_.]?2[-_.]?0|seed[-_.]?2[-_.]?0|qwen[-_.]?vl|glm[-_.]?4v|gpt[-_.]?4o|vision|vl|多模态)/i.test(
+      modelText,
+    )
+  ) {
+    return true;
+  }
+
+  return false;
+}
+
 function pickQwenVisionModel(models: EnabledModelWithProvider[]) {
-  const visionModels = models.filter((item) => item.model.supportsVision && item.model.isEnabled);
+  const visionModels = models.filter((item) => isLikelyVisionCapableModel(item));
   const zhipuVision = visionModels.find(
     (item) =>
       item.provider.type === "zhipu" ||
@@ -514,12 +550,7 @@ function pickAgentVisionModel(
   agentModel: EnabledModelWithProvider | undefined,
   models: EnabledModelWithProvider[],
 ) {
-  if (
-    agentModel &&
-    agentModel.model.isEnabled &&
-    agentModel.model.supportsVision &&
-    isAiModelMultimodal(agentModel.model)
-  ) {
+  if (agentModel && isLikelyVisionCapableModel(agentModel)) {
     return agentModel;
   }
 
diff --git a/lib/ai/smart-analyzer.ts b/lib/ai/smart-analyzer.ts
index 0698cd557..91fa20832 100644
--- a/lib/ai/smart-analyzer.ts
+++ b/lib/ai/smart-analyzer.ts
@@ -568,6 +568,56 @@ interface HistoryMessage {
   content: string;
 }
 
+function isLikelyVisionCapableModel(modelWithProvider: {
+  model: {
+    isEnabled: boolean;
+    isSubModel: boolean;
+    supportsVision?: boolean;
+    displayName?: string | null;
+    modelId?: string | null;
+    capabilities?: unknown;
+    defaultParams?: unknown;
+  };
+  provider: {
+    type?: string | null;
+  };
+}) {
+  const model = modelWithProvider.model;
+  const providerType = String(modelWithProvider.provider?.type || "");
+  if (!model?.isEnabled || model.isSubModel) return false;
+  if (model.supportsVision) return true;
+  if (isAiModelMultimodal(model)) return true;
+
+  const modelText = `${model.displayName || ""} ${model.modelId || ""}`.toLowerCase();
+
+  if (providerType === "volcano-multimodal") return true;
+
+  if (
+    providerType === "zhipu" &&
+    /(glm[-_.]?4v|glm[-_.]?4\.?6v|vision|多模态)/i.test(modelText)
+  ) {
+    return true;
+  }
+
+  if (
+    providerType === "alibaba" &&
+    /(qwen|vl|vision|通义|多模态)/i.test(modelText)
+  ) {
+    return true;
+  }
+
+  if (
+    ["custom", "openai", "local"].includes(providerType) &&
+    /(doubao[-_.]?seed[-_.]?2[-_.]?0|seed[-_.]?2[-_.]?0|qwen[-_.]?vl|glm[-_.]?4v|gpt[-_.]?4o|vision|vl|多模态)/i.test(
+      modelText,
+    )
+  ) {
+    return true;
+  }
+
+  return false;
+}
+
 /**
  * 分析用户意图
  */
@@ -657,17 +707,13 @@ async function analyzWithLLM(
     );
   const zhipuVisionModel = allModels.find(
     (m) =>
-      m.model.supportsVision &&
-      m.model.isEnabled &&
-      !m.model.isSubModel &&
+      isLikelyVisionCapableModel(m) &&
       isLikelyZhipuVisionModel(m.model.displayName, m.model.modelId, m.provider.type),
   );
   const qwenVisionModel = allModels.find(
     (m) =>
       m.provider.type === "alibaba" &&
-      m.model.supportsVision &&
-      m.model.isEnabled &&
-      !m.model.isSubModel &&
+      isLikelyVisionCapableModel(m) &&
       /(qwen|通义|vl)/i.test(`${m.model.displayName} ${m.model.modelId}`)
   );
   const fallbackVisionModel =
@@ -675,25 +721,23 @@ async function analyzWithLLM(
     allModels.find(
       (m) =>
         m.provider.type === "alibaba" &&
-        m.model.supportsVision &&
-        m.model.isEnabled &&
-        !m.model.isSubModel,
+        isLikelyVisionCapableModel(m),
     );
-  const preferredVisionModel = zhipuVisionModel || fallbackVisionModel;
+  const preferredVisionModel =
+    zhipuVisionModel ||
+    fallbackVisionModel ||
+    allModels.find((m) => isLikelyVisionCapableModel(m));
   const usableAgentModel =
     agentModel && isReasoningCompatibleProvider(agentModel.provider.type)
       ? agentModel
       : undefined;
   const multimodalAgentVisionModel =
     agentModel &&
-    agentModel.model.isEnabled &&
-    !agentModel.model.isSubModel &&
-    agentModel.model.supportsVision &&
-    isAiModelMultimodal(agentModel.model)
+    isLikelyVisionCapableModel(agentModel)
       ? agentModel
       : undefined;
   const selectedAnalyzerModel = hasImage
-    ? multimodalAgentVisionModel || preferredVisionModel
+    ? multimodalAgentVisionModel || preferredVisionModel || usableAgentModel
     : usableAgentModel ||
       allModels.find((m) => m.provider.type === "zhipu" && !m.model.isSubModel) ||
       preferredVisionModel;