模块介绍 / 今日规划 / 下班日报 / 团队摘要 生成成功后没有上报，也几乎不写「AI」日志。

流结束时的 usage 用了非阻塞发送，通道一满就被丢掉；很多服务商默认还不在流里带 usage。
现在每次真正打到模型的调用都会：

在本地日志（分类 AI，消息 AI 调用完成）写下 scene / prompt / completion / cached
登录后同步上报到后台日汇总

service/ai/ai_test.go

@@ -56,6 +56,51 @@ func TestChatStreamSSE(t *testing.T) {
 	}
 }
 
+func TestParseUsageCachedTokens(t *testing.T) {
+	u := parseUsage(&chatUsage{
+		PromptTokens:         100,
+		CompletionTokens:     20,
+		PromptCacheHitTokens: 80,
+	})
+	if u == nil || u.CachedTokens != 80 || u.PromptTokens != 100 {
+		t.Fatalf("deepseek cache field: %+v", u)
+	}
+	u = parseUsage(&chatUsage{
+		PromptTokens:     90,
+		CompletionTokens: 10,
+		PromptTokensDetails: &struct {
+			CachedTokens int64 `json:"cached_tokens"`
+		}{CachedTokens: 64},
+	})
+	if u == nil || u.CachedTokens != 64 {
+		t.Fatalf("openai cached_tokens: %+v", u)
+	}
+}
+
+func TestChatStreamSendsUsage(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "text/event-stream")
+		w.Write([]byte("data: {\"choices\":[{\"delta\":{\"content\":\"ok\"}}]}\n\n"))
+		w.Write([]byte("data: {\"usage\":{\"prompt_tokens\":12,\"completion_tokens\":3,\"prompt_cache_hit_tokens\":8}}\n\n"))
+		w.Write([]byte("data: [DONE]\n\n"))
+	}))
+	defer srv.Close()
+	p := &openAICompatible{name: "test", baseURL: srv.URL, model: "m", apiKey: "k", client: srv.Client()}
+	stream, e := p.ChatStream(context.Background(), []Message{{Role: "user", Content: "hi"}})
+	if e != nil {
+		t.Fatal(e)
+	}
+	var usage *Usage
+	for c := range stream {
+		if c.Usage != nil {
+			usage = c.Usage
+		}
+	}
+	if usage == nil || usage.PromptTokens != 12 || usage.CachedTokens != 8 {
+		t.Fatalf("usage not delivered: %+v", usage)
+	}
+}
+
 func TestChatStreamAuthError(t *testing.T) {
 	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(http.StatusUnauthorized)