fix(llm): 移除测试用 token 限制，保留完整的推理模型 temperature 兼容与 Usage 空指针修复

internal/llm/client.go

@@ -238,16 +238,50 @@ func NewClient(cfg Config) (*Client, error) {
 	}, nil
 }
 
+// isReasoningModel 判断是否为推理或带有 Beta 限制的模型（这类模型要求 temperature 固定为 1）
+func isReasoningModel(model string) bool {
+	m := strings.ToLower(model)
+	return strings.HasPrefix(m, "o1") ||
+		strings.HasPrefix(m, "o3") ||
+		strings.HasPrefix(m, "o4") ||
+		strings.Contains(m, "reasoner") ||
+		strings.Contains(m, "gpt-5") ||
+		strings.Contains(m, "luna") ||
+		strings.Contains(m, "terra") ||
+		strings.Contains(m, "sol")
+}
+
+// isTemperatureError 检测 API 报错是否由 temperature 参数限制触发
+func isTemperatureError(err error) bool {
+	if err == nil {
+		return false
+	}
+	msg := strings.ToLower(err.Error())
+	return strings.Contains(msg, "temperature") || strings.Contains(msg, "beta-limitations")
+}
+
 // Review 一次性完整返回代码评审结果
 func (c *Client) Review(ctx context.Context, systemPrompt, diff string) (string, error) {
-	resp, err := c.client.CreateChatCompletion(ctx, openai.ChatCompletionRequest{
+	temp := float32(0.2)
+	if isReasoningModel(c.model) {
+		temp = 1.0
+	}
+
+	req := openai.ChatCompletionRequest{
 		Model: c.model,
 		Messages: []openai.ChatCompletionMessage{
 			{Role: openai.ChatMessageRoleSystem, Content: systemPrompt},
 			{Role: openai.ChatMessageRoleUser, Content: diff},
 		},
-		Temperature: 0.2, // 代码评审低随机性
+		Temperature: temp,
-	})
+	}
+
+	resp, err := c.client.CreateChatCompletion(ctx, req)
+	// 针对限制了 temperature 的模型，且首次非 1.0 时，自动调整为 1.0 重试
+	if err != nil && req.Temperature != 1.0 && isTemperatureError(err) {
+		req.Temperature = 1.0
+		resp, err = c.client.CreateChatCompletion(ctx, req)
+	}
 
 	if err != nil {
 		return "", fmt.Errorf("代码评审调用失败: %w", err)
@@ -376,18 +410,31 @@ func (u *Usage) EstimateCost() float64 {
 
 // ReviewStream 流式逐字返回评审结果（打字机效果），同时返回 Token 使用量
 func (c *Client) ReviewStream(ctx context.Context, systemPrompt, diff string, onChunk func(chunk string)) (*Usage, error) {
-	stream, err := c.client.CreateChatCompletionStream(ctx, openai.ChatCompletionRequest{
+	temp := float32(0.2)
+	if isReasoningModel(c.model) {
+		temp = 1.0
+	}
+
+	req := openai.ChatCompletionRequest{
 		Model: c.model,
 		Messages: []openai.ChatCompletionMessage{
 			{Role: openai.ChatMessageRoleSystem, Content: systemPrompt},
 			{Role: openai.ChatMessageRoleUser, Content: diff},
 		},
-		Temperature: 0.2,
+		Temperature: temp,
 		Stream:      true,
 		StreamOptions: &openai.StreamOptions{
 			IncludeUsage: true,
 		},
-	})
+	}
+
+	stream, err := c.client.CreateChatCompletionStream(ctx, req)
+	// 针对限制了 temperature 的模型，且首次非 1.0 时，自动调整为 1.0 重试
+	if err != nil && req.Temperature != 1.0 && isTemperatureError(err) {
+		req.Temperature = 1.0
+		stream, err = c.client.CreateChatCompletionStream(ctx, req)
+	}
+
 	if err != nil {
 		return nil, fmt.Errorf("创建流式请求失败: %w", err)
 	}
@@ -417,7 +464,7 @@ func (c *Client) ReviewStream(ctx context.Context, systemPrompt, diff string, on
 		}
 
 		// 捕获最后一个 chunk 中携带的 Usage 信息
-		if response.Usage.TotalTokens > 0 {
+		if response.Usage != nil && response.Usage.TotalTokens > 0 {
 			usage.InputTokens = response.Usage.PromptTokens
 			usage.OutputTokens = response.Usage.CompletionTokens
 		}

internal/llm/model_test.go

@@ -0,0 +1,51 @@
+package llm
+
+import (
+	"fmt"
+	"testing"
+)
+
+func TestIsReasoningModel(t *testing.T) {
+	tests := []struct {
+		model    string
+		expected bool
+	}{
+		{"o1", true},
+		{"o1-mini", true},
+		{"o1-preview", true},
+		{"o3-mini", true},
+		{"gpt-5.6-luna", true},
+		{"gpt-5.6-sol", true},
+		{"gpt-5.6-terra", true},
+		{"deepseek-reasoner", true},
+		{"gemini-3.7-flash", false},
+		{"claude-3.7-sonnet", false},
+		{"gpt-4o", false},
+	}
+
+	for _, tt := range tests {
+		got := isReasoningModel(tt.model)
+		if got != tt.expected {
+			t.Errorf("isReasoningModel(%q) = %v, expected %v", tt.model, got, tt.expected)
+		}
+	}
+}
+
+func TestIsTemperatureError(t *testing.T) {
+	tests := []struct {
+		errStr   string
+		expected bool
+	}{
+		{"this model has beta-limitations, temperature, top_p and n are fixed at 1", true},
+		{"temperature must be 1.0", true},
+		{"invalid api key", false},
+		{"rate limit exceeded", false},
+	}
+
+	for _, tt := range tests {
+		got := isTemperatureError(fmt.Errorf("%s", tt.errStr))
+		if got != tt.expected {
+			t.Errorf("isTemperatureError(%q) = %v, expected %v", tt.errStr, got, tt.expected)
+		}
+	}
+}

internal/updater/checker.go

@@ -12,7 +12,7 @@ import (
 )
 
 // CurrentVersion 当前编译版本号
-const CurrentVersion = "v1.0.5"
+const CurrentVersion = "v1.0.6"
 
 const (
 	githubAPIURL  = "https://api.github.com/repos/unclesam-ly/git-air/releases/latest"