Add token usage stats and consolidate credentials in SQLite

- New internal/stats (types) and internal/store (SQLite owner: requests + credentials tables, WAL); store implements stats.Recorder.
- Stream (SSE tee) and non-stream chat paths parse upstream usage incl. cached_tokens and record per-request; add /api/stats, /api/stats/reset, /admin/stats HTML with cache hit rate.
- Drop credentials.json: remove auth file I/O and ZHANLU_CREDENTIALS_FILE; credential precedence is env vars > db row.

internal/server/server_test.go

@@ -15,12 +15,15 @@ import (
 
 	"git.misaka.ren/M1saka/zhanlu_proxy/internal/auth"
 	"git.misaka.ren/M1saka/zhanlu_proxy/internal/config"
+	"git.misaka.ren/M1saka/zhanlu_proxy/internal/store"
 )
 
 const testSM2Key = "8d6ee90b3c4d299ae5abd655dbc3547c110ae8aeff1de18b0df241f215f90748"
 
-// setupTestServer spins up a mock Zhanlu upstream and a proxy server wired to it.
-func setupTestServer(t *testing.T) (*httptest.Server, *httptest.Server, string) {
+// setupTestServer spins up a mock Zhanlu upstream and a proxy server wired to
+// it. The proxy is backed by a temp SQLite store so credentials persist in the
+// db during the test instead of a JSON file.
+func setupTestServer(t *testing.T) (*httptest.Server, *httptest.Server, *store.Store) {
 	t.Helper()
 	upstream := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		switch r.URL.Path {
@@ -47,7 +50,7 @@ func setupTestServer(t *testing.T) (*httptest.Server, *httptest.Server, string)
 				return
 			}
 			w.Header().Set("Content-Type", "text/event-stream")
-			_, _ = fmt.Fprint(w, "data: {\"id\":\"chatcmpl-x\",\"choices\":[{\"index\":0,\"delta\":{\"content\":\"hello\"},\"finish_reason\":null}],\"usage\":{\"prompt_tokens\":1}}\n\n")
+			_, _ = fmt.Fprint(w, "data: {\"id\":\"chatcmpl-x\",\"choices\":[{\"index\":0,\"delta\":{\"content\":\"hello\"},\"finish_reason\":null}],\"usage\":{\"prompt_tokens\":10,\"completion_tokens\":20,\"total_tokens\":30,\"prompt_tokens_details\":{\"cached_tokens\":4}}}\n\n")
 			_, _ = fmt.Fprint(w, "data: {\"id\":\"chatcmpl-x\",\"choices\":[{\"index\":0,\"delta\":{},\"finish_reason\":\"stop\"}]}\n\n")
 			_, _ = fmt.Fprint(w, "data: [DONE]\n\n")
 		case "/gateway/v1/model/info":
@@ -59,28 +62,32 @@ func setupTestServer(t *testing.T) (*httptest.Server, *httptest.Server, string)
 		}
 	}))
 
-	credsFile := filepath.Join(t.TempDir(), "credentials.json")
+	st, err := store.Open(filepath.Join(t.TempDir(), "stats.db"))
+	if err != nil {
+		t.Fatalf("open store: %v", err)
+	}
+	t.Cleanup(func() { st.Close() })
 	cfg := config.Config{
 		ListenAddr:         ":0",
 		MobileLoginBaseURL: upstream.URL,
 		MobileModelBaseURL: upstream.URL,
 		UpstreamPath:       "/chat/completions",
-		CredentialsPath:    credsFile,
+		DBPath:             filepath.Join(t.TempDir(), "zhanlu.db"),
 		TokenDecryptKey:    "3jw7woww2rvhla6k",
 		PublicKeyPEM:       defaultTestPublicKey,
 		PhonePublicKeyPEM:  defaultTestPublicKey,
 		SM2PrivateKey:      testSM2Key,
 		PluginVersion:      "1.4.2",
 	}
-	h := New(cfg)
+	h := New(cfg, st)
 	proxy := httptest.NewServer(h)
-	return upstream, proxy, credsFile
+	return upstream, proxy, st
 }
 
 // TestPhoneLoginAndChat exercises the full v1.4.2 flow: SMS login, profile
 // fetch, SM2 API-key provisioning, then OpenAI-compatible chat and models.
 func TestPhoneLoginAndChat(t *testing.T) {
-	upstream, proxy, credsFile := setupTestServer(t)
+	upstream, proxy, st := setupTestServer(t)
 	defer upstream.Close()
 	defer proxy.Close()
 
@@ -107,8 +114,8 @@ func TestPhoneLoginAndChat(t *testing.T) {
 		t.Fatalf("login failed: %v", loginResp)
 	}
 
-	// 3. credentials file should contain the provisioned api key
-	creds, err := auth.LoadCredentials(credsFile)
+	// 3. credentials store should contain the provisioned api key
+	creds, err := st.LoadCredentials()
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -154,11 +161,11 @@ func TestPhoneLoginAndChat(t *testing.T) {
 }
 
 func TestStreamingChat(t *testing.T) {
-	upstream, proxy, credsFile := setupTestServer(t)
+	upstream, proxy, st := setupTestServer(t)
 	defer upstream.Close()
 	defer proxy.Close()
 
-	// Seed credentials directly with the api key
+	// Seed credentials directly with the api key into the store
 	creds := auth.Credentials{
 		AccessKey:    "AK",
 		SecretKey:    "SK",
@@ -167,7 +174,7 @@ func TestStreamingChat(t *testing.T) {
 		ModelBaseURL: upstream.URL,
 		Email:        "[email protected]",
 	}
-	if err := auth.SaveCredentials(credsFile, creds); err != nil {
+	if err := st.SaveCredentials(creds); err != nil {
 		t.Fatal(err)
 	}
 
@@ -188,6 +195,88 @@ func TestStreamingChat(t *testing.T) {
 	}
 }
 
+// TestStats verifies both streaming and non-streaming chat paths record token
+// usage into the store and that GET /api/stats aggregates them correctly.
+func TestStats(t *testing.T) {
+	upstream, proxy, st := setupTestServer(t)
+	defer upstream.Close()
+	defer proxy.Close()
+
+	creds := auth.Credentials{
+		AccessKey: "AK", SecretKey: "SK", Token: "TOKEN",
+		APIKey: "sk-test-456", ModelBaseURL: upstream.URL, Email: "[email protected]",
+	}
+	if err := st.SaveCredentials(creds); err != nil {
+		t.Fatal(err)
+	}
+
+	// non-streaming chat
+	resp, err := http.Post(proxy.URL+"/v1/chat/completions", "application/json", strings.NewReader(`{"model":"GLM-4.7","messages":[{"role":"user","content":"hi"}],"stream":false}`))
+	if err != nil {
+		t.Fatal(err)
+	}
+	io.Copy(io.Discard, resp.Body)
+	resp.Body.Close()
+
+	// streaming chat
+	resp, err = http.Post(proxy.URL+"/v1/chat/completions", "application/json", strings.NewReader(`{"model":"GLM-4.7","messages":[{"role":"user","content":"hi"}],"stream":true}`))
+	if err != nil {
+		t.Fatal(err)
+	}
+	io.Copy(io.Discard, resp.Body)
+	resp.Body.Close()
+
+	// query stats
+	resp, err = http.Get(proxy.URL + "/api/stats")
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer resp.Body.Close()
+	var statsResp map[string]any
+	if err := json.NewDecoder(resp.Body).Decode(&statsResp); err != nil {
+		t.Fatal(err)
+	}
+	if enabled, _ := statsResp["enabled"].(bool); !enabled {
+		t.Fatalf("stats not enabled: %v", statsResp)
+	}
+	s, _ := statsResp["stats"].(map[string]any)
+	if s == nil {
+		t.Fatalf("no stats object: %v", statsResp)
+	}
+	totals, _ := s["totals"].(map[string]any)
+	if totals == nil {
+		t.Fatalf("no totals: %v", s)
+	}
+	if got := totNum(totals["requests"]); got != 2 {
+		t.Fatalf("requests = %v, want 2", totals["requests"])
+	}
+	if got := totNum(totals["prompt_tokens"]); got != 20 {
+		t.Fatalf("prompt_tokens = %v, want 20", totals["prompt_tokens"])
+	}
+	if got := totNum(totals["completion_tokens"]); got != 40 {
+		t.Fatalf("completion_tokens = %v, want 40", totals["completion_tokens"])
+	}
+	if got := totNum(totals["total_tokens"]); got != 60 {
+		t.Fatalf("total_tokens = %v, want 60", totals["total_tokens"])
+	}
+	if got := totNum(totals["cached_tokens"]); got != 8 {
+		t.Fatalf("cached_tokens = %v, want 8", totals["cached_tokens"])
+	}
+	if rate, _ := totals["cache_rate"].(float64); rate < 0.39 || rate > 0.41 {
+		t.Fatalf("cache_rate = %v, want ~0.4", totals["cache_rate"])
+	}
+	perModel, _ := s["per_model"].([]any)
+	if len(perModel) != 1 {
+		t.Fatalf("per_model = %v, want 1 entry", perModel)
+	}
+}
+
+// totNum extracts an int from a JSON-decoded numeric value (float64).
+func totNum(v any) int {
+	f, _ := v.(float64)
+	return int(f)
+}
+
 func okValue(m map[string]any) bool {
 	ok, _ := m["ok"].(bool)
 	return ok