diff --git a/crates/zclaw-runtime/src/test_util.rs b/crates/zclaw-runtime/src/test_util.rs
index 7f01330..25e6647 100644
--- a/crates/zclaw-runtime/src/test_util.rs
+++ b/crates/zclaw-runtime/src/test_util.rs
@@ -79,7 +79,7 @@ impl MockLlmDriver {
     }
 
     /// Queue an error response.
-    pub fn with_error(self, _error: &str) -> Self {
+    pub fn with_error(mut self, _error: &str) -> Self {
         self.push_response(CompletionResponse {
             content: vec![],
             model: "mock-model".to_string(),
@@ -97,7 +97,7 @@ impl MockLlmDriver {
     }
 
     /// Queue stream chunks for a streaming call.
-    pub fn with_stream_chunks(mut self, chunks: Vec<StreamChunk>) -> Self {
+    pub fn with_stream_chunks(self, chunks: Vec<StreamChunk>) -> Self {
         self.stream_chunks
             .lock()
             .expect("stream_chunks lock")
diff --git a/crates/zclaw-saas/src/lib.rs b/crates/zclaw-saas/src/lib.rs
index d01b4da..528416e 100644
--- a/crates/zclaw-saas/src/lib.rs
+++ b/crates/zclaw-saas/src/lib.rs
@@ -28,3 +28,5 @@ pub mod telemetry;
 pub mod billing;
 pub mod industry;
 pub mod knowledge;
+
+pub use error::{SaasError, SaasError as Error, SaasResult as Result};
diff --git a/desktop/src/components/ChatArea.tsx b/desktop/src/components/ChatArea.tsx
index 80caafa..8aeb296 100644
--- a/desktop/src/components/ChatArea.tsx
+++ b/desktop/src/components/ChatArea.tsx
@@ -88,12 +88,17 @@ export function ChatArea({ compact, onOpenDetail }: { compact?: boolean; onOpenD
   const models = useMemo(() => {
     const failed = failedModelIds.current;
     if (isLoggedIn && saasModels.length > 0) {
-      return saasModels.map(m => ({
-        id: m.alias || m.id,
-        name: m.alias || m.id,
-        provider: m.provider_id,
-        available: !failed.has(m.alias || m.id),
-      }));
+      return saasModels
+        .filter(m => {
+          const name = (m.alias || m.id).toLowerCase();
+          return !name.includes('embedding');
+        })
+        .map(m => ({
+          id: m.alias || m.id,
+          name: m.alias || m.id,
+          provider: undefined,
+          available: !failed.has(m.alias || m.id),
+        }));
     }
     if (configModels.length > 0) {
       return configModels;
@@ -210,6 +215,8 @@ export function ChatArea({ compact, onOpenDetail }: { compact?: boolean; onOpenD
       'hand-execution-complete',
       (event) => {
         const { handId, success, error } = event.payload;
+        const streaming = useChatStore.getState().isStreaming;
+        if (!streaming) return;
         useChatStore.getState().addMessage({
           id: crypto.randomUUID(),
           role: 'hand',
@@ -502,7 +509,7 @@ export function ChatArea({ compact, onOpenDetail }: { compact?: boolean; onOpenD
           {!isStreaming && suggestions.length > 0 && !messages.some(m => m.error) && (
             <SuggestionChips
               suggestions={suggestions}
-              onSelect={(text) => { setInput(text); textareaRef.current?.focus(); }}
+              onSelect={(text) => { setInput(text); textareaRef.current?.focus(); setTimeout(() => handleSend(), 0); }}
               className="mb-3"
             />
           )}
diff --git a/desktop/src/store/chat/streamStore.ts b/desktop/src/store/chat/streamStore.ts
index c0aebfc..9eb178a 100644
--- a/desktop/src/store/chat/streamStore.ts
+++ b/desktop/src/store/chat/streamStore.ts
@@ -37,6 +37,39 @@ import { useArtifactStore } from './artifactStore';
 
 const log = createLogger('StreamStore');
 
+// ---------------------------------------------------------------------------
+// Error formatting — convert raw LLM/API errors to user-friendly messages
+// ---------------------------------------------------------------------------
+
+function formatUserError(raw: string): string {
+  if (raw.includes('API Key') || raw.includes('没有可用的 API Key')) {
+    return '模型服务暂时不可用，请稍后重试';
+  }
+  if (raw.includes('404') || raw.includes('NOT_FOUND')) {
+    return '模型服务未找到，请检查模型配置';
+  }
+  if (raw.includes('429') || raw.includes('rate_limit') || raw.includes('Too Many Requests')) {
+    return '请求过于频繁，请稍后重试';
+  }
+  if (raw.includes('401') || raw.includes('Unauthorized')) {
+    return '认证已过期，请重新登录';
+  }
+  if (raw.includes('402') || raw.includes('quota') || raw.includes('配额')) {
+    return '使用配额已用尽，请升级订阅或联系管理员';
+  }
+  if (raw.includes('timeout') || raw.includes('超时') || raw.includes('Timeout')) {
+    return '请求超时，请稍后重试';
+  }
+  if (raw.includes('502') || raw.includes('Bad Gateway')) {
+    return '服务网关异常，请稍后重试';
+  }
+  if (raw.includes('503') || raw.includes('Service Unavailable')) {
+    return '服务暂时不可用，请稍后重试';
+  }
+  // Strip raw JSON from remaining errors
+  return raw.replace(/\{[^}]*\}/g, '').replace(/\s+/g, ' ').trim().substring(0, 80) || '请求失败，请稍后重试';
+}
+
 // ---------------------------------------------------------------------------
 // 401 Auth Error Recovery
 // ---------------------------------------------------------------------------
@@ -556,7 +589,8 @@ export const useStreamStore = create<StreamState>()(
 
             // Attempt 401 auth recovery (token refresh + kernel reconnect)
             const recoveryMsg = await tryRecoverFromAuthError(error);
-            const displayError = recoveryMsg || error;
+            const rawError = recoveryMsg || error;
+            const displayError = formatUserError(rawError);
 
             _chat?.updateMessages(msgs =>
               msgs.map(m =>
@@ -699,6 +733,9 @@ export const useStreamStore = create<StreamState>()(
     }
 
     const unsubscribe = client.onAgentStream((delta: AgentStreamDelta) => {
+      const activeRunId = get().activeRunId;
+      if (activeRunId && delta.runId && delta.runId !== activeRunId) return;
+
       const msgs = _chat?.getMessages() || [];
 
       const streamingMsg = [...msgs]
@@ -710,10 +747,7 @@ export const useStreamStore = create<StreamState>()(
             (delta.runId && m.runId === delta.runId)
             || (!delta.runId && m.runId === null)
           )
-        ))
-        || [...msgs]
-          .reverse()
-          .find(m => m.role === 'assistant' && m.streaming);
+        ));
 
       if (!streamingMsg) return;
 
@@ -804,6 +838,8 @@ export const useStreamStore = create<StreamState>()(
           }
         }
       } else if (delta.stream === 'hand') {
+        const runId = get().activeRunId;
+        if (!runId || (delta.runId && delta.runId !== runId)) return;
         const handMsg: StreamMsg = {
           id: `hand_${Date.now()}_${generateRandomString(4)}`,
           role: 'hand',
@@ -818,6 +854,8 @@ export const useStreamStore = create<StreamState>()(
         };
         _chat?.updateMessages(ms => [...ms, handMsg]);
       } else if (delta.stream === 'workflow') {
+        const runId = get().activeRunId;
+        if (!runId || (delta.runId && delta.runId !== runId)) return;
         const workflowMsg: StreamMsg = {
           id: `workflow_${Date.now()}_${generateRandomString(4)}`,
           role: 'workflow',
diff --git a/desktop/src/store/chatStore.ts b/desktop/src/store/chatStore.ts
index 8a7c10c..fba6047 100644
--- a/desktop/src/store/chatStore.ts
+++ b/desktop/src/store/chatStore.ts
@@ -155,12 +155,18 @@ export const useChatStore = create<ChatState>()(
   },
 
   newConversation: () => {
+    if (get().isStreaming) {
+      useStreamStore.getState().cancelStream();
+    }
     const messages = get().messages;
     useConversationStore.getState().newConversation(messages);
     set({ messages: [], isStreaming: false });
   },
 
   switchConversation: (id: string) => {
+    if (get().isStreaming) {
+      useStreamStore.getState().cancelStream();
+    }
     const messages = get().messages;
     const result = useConversationStore.getState().switchConversation(id, messages);
     if (result) {