feat: optimize WebUI stream output and sanitize user-facing answers

internal/llm/client.go

@@ -81,9 +81,10 @@ type InputFile struct {
 }
 
 type OpenAICompatibleClient struct {
-	client openai.Client
-	model  string
-	log    *logger.Logger
+	client               openai.Client
+	model                string
+	disableThinkingParam bool
+	log                  *logger.Logger
 }
 
 func NewOpenAICompatibleClient(cfg config.LLMConfig, log *logger.Logger) *OpenAICompatibleClient {
@@ -95,9 +96,10 @@ func NewOpenAICompatibleClient(cfg config.LLMConfig, log *logger.Logger) *OpenAI
 		opts = append(opts, option.WithBaseURL(cfg.BaseURL))
 	}
 	return &OpenAICompatibleClient{
-		client: openai.NewClient(opts...),
-		model:  cfg.Model,
-		log:    log,
+		client:               openai.NewClient(opts...),
+		model:                cfg.Model,
+		disableThinkingParam: shouldDisableThinkingParam(cfg.BaseURL),
+		log:                  log,
 	}
 }
 
@@ -138,7 +140,7 @@ func (c *OpenAICompatibleClient) GenerateWithTools(ctx context.Context, messages
 		}
 	}
 
-	resp, err := c.client.Chat.Completions.New(ctx, params)
+	resp, err := c.client.Chat.Completions.New(ctx, params, c.chatCompletionRequestOptions()...)
 	if err != nil {
 		return nil, fmt.Errorf("llm tool-call request failed: %w", err)
 	}
@@ -180,7 +182,7 @@ func (c *OpenAICompatibleClient) generateWithMessagesInternal(ctx context.Contex
 		Messages: sdkMessages,
 	}
 
-	resp, err := c.client.Chat.Completions.New(ctx, params)
+	resp, err := c.client.Chat.Completions.New(ctx, params, c.chatCompletionRequestOptions()...)
 	if err != nil {
 		if c.log != nil {
 			c.log.Errorf("llm request failed err=%v", err)
@@ -392,3 +394,18 @@ func appendIfMissing(items []string, value string) []string {
 	}
 	return append(items, value)
 }
+
+func (c *OpenAICompatibleClient) chatCompletionRequestOptions() []option.RequestOption {
+	if !c.disableThinkingParam {
+		return nil
+	}
+	return []option.RequestOption{option.WithJSONSet("enable_thinking", false)}
+}
+
+func shouldDisableThinkingParam(baseURL string) bool {
+	baseURL = strings.ToLower(strings.TrimSpace(baseURL))
+	if baseURL == "" {
+		return false
+	}
+	return strings.Contains(baseURL, "dashscope.aliyuncs.com")
+}

internal/llm/client_test.go

@@ -0,0 +1,24 @@
+package llm
+
+import "testing"
+
+func TestShouldDisableThinkingParam(t *testing.T) {
+	if !shouldDisableThinkingParam("https://dashscope.aliyuncs.com/compatible-mode/v1") {
+		t.Fatal("expected DashScope base URL to require enable_thinking=false")
+	}
+	if shouldDisableThinkingParam("https://api.openai.com/v1") {
+		t.Fatal("expected standard OpenAI base URL not to require enable_thinking=false")
+	}
+}
+
+func TestChatCompletionRequestOptions(t *testing.T) {
+	client := &OpenAICompatibleClient{disableThinkingParam: true}
+	if got := len(client.chatCompletionRequestOptions()); got != 1 {
+		t.Fatalf("expected 1 request option when disableThinkingParam=true, got %d", got)
+	}
+
+	client.disableThinkingParam = false
+	if got := len(client.chatCompletionRequestOptions()); got != 0 {
+		t.Fatalf("expected 0 request options when disableThinkingParam=false, got %d", got)
+	}
+}