fix(grok): estimate messages count tokens locally

backend/internal/server/routes/gateway.go

@@ -43,6 +43,16 @@ func RegisterGatewayRoutes(
 	isOpenAIGatewayPlatform := func(c *gin.Context) bool {
 		return getGroupPlatform(c) == service.PlatformOpenAI
 	}
+	countTokensHandler := func(c *gin.Context) {
+		switch getGroupPlatform(c) {
+		case service.PlatformOpenAI:
+			h.OpenAIGateway.CountTokens(c)
+		case service.PlatformGrok:
+			h.OpenAIGateway.GrokCountTokens(c)
+		default:
+			h.Gateway.CountTokens(c)
+		}
+	}
 	modelsHandler := func(c *gin.Context) {
 		if isOpenAIGatewayPlatform(c) && c.Query("client_version") != "" {
 			h.OpenAIGateway.CodexModels(c)
@@ -139,26 +149,9 @@ func RegisterGatewayRoutes(
 			}
 			h.Gateway.Messages(c)
 		})
-		// /v1/messages/count_tokens: OpenAI uses Anthropic-compat bridge; other
-		// OpenAI-compatible platforms keep the prior unsupported response.
-		gateway.POST("/messages/count_tokens", func(c *gin.Context) {
-			if isOpenAIGatewayPlatform(c) {
-				h.OpenAIGateway.CountTokens(c)
-				return
-			}
-			if isOpenAIResponsesCompatibleGatewayPlatform(c) {
-				service.MarkOpsClientBusinessLimited(c, service.OpsClientBusinessLimitedReasonLocalFeatureGate)
-				c.JSON(http.StatusNotFound, gin.H{
-					"type": "error",
-					"error": gin.H{
-						"type":    "not_found_error",
-						"message": "Token counting is not supported for this platform",
-					},
-				})
-				return
-			}
-			h.Gateway.CountTokens(c)
-		})
+		// /v1/messages/count_tokens: OpenAI bridges upstream, Grok estimates
+		// locally, and Anthropic-compatible platforms retain their existing path.
+		gateway.POST("/messages/count_tokens", countTokensHandler)
 		// Codex CLI / Codex app refresh their model picker from the provider's
 		// /models endpoint with a client_version query and expect the ChatGPT
 		// Codex manifest format; other clients keep the OpenAI-style list.
@@ -256,6 +249,7 @@ func RegisterGatewayRoutes(
 		h.OpenAIGateway.ResponsesWebSocket(c)
 	})
 	r.GET("/models", bodyLimit, clientRequestID, opsErrorLogger, endpointNorm, gin.HandlerFunc(apiKeyAuth), requireGroupAnthropic, modelsHandler)
+	r.POST("/messages/count_tokens", bodyLimit, clientRequestID, opsErrorLogger, endpointNorm, gin.HandlerFunc(apiKeyAuth), requireGroupAnthropic, countTokensHandler)
 	codexDirect := r.Group("/backend-api/codex")
 	codexDirect.Use(bodyLimit, clientRequestID, opsErrorLogger, endpointNorm, gin.HandlerFunc(apiKeyAuth), requireGroupAnthropic)
 	{

backend/internal/server/routes/gateway_test.go

@@ -1,6 +1,7 @@
 package routes
 
 import (
+	"encoding/json"
 	"net/http"
 	"net/http/httptest"
 	"strings"
@@ -15,6 +16,10 @@ import (
 )
 
 func newGatewayRoutesTestRouter(platform ...string) *gin.Engine {
+	return newGatewayRoutesTestRouterWithConfig(&config.Config{}, platform...)
+}
+
+func newGatewayRoutesTestRouterWithConfig(cfg *config.Config, platform ...string) *gin.Engine {
 	gin.SetMode(gin.TestMode)
 	router := gin.New()
 
@@ -42,7 +47,7 @@ func newGatewayRoutesTestRouter(platform ...string) *gin.Engine {
 		nil,
 		nil,
 		nil,
-		&config.Config{},
+		cfg,
 	)
 
 	return router
@@ -224,13 +229,22 @@ func TestGatewayRoutesGrokAllowsCLICompatibilityEntrypoints(t *testing.T) {
 		require.NotContains(t, w.Body.String(), "not supported for Grok groups")
 	}
 
-	req := httptest.NewRequest(http.MethodPost, "/v1/messages/count_tokens", strings.NewReader(`{"model":"grok","messages":[{"role":"user","content":"hi"}]}`))
-	req.Header.Set("Content-Type", "application/json")
-	w := httptest.NewRecorder()
+	countTokensRouter := newGatewayRoutesTestRouterWithConfig(&config.Config{
+		Gateway: config.GatewayConfig{MaxBodySize: 1024 * 1024},
+	}, service.PlatformGrok)
+	for _, path := range []string{"/v1/messages/count_tokens", "/messages/count_tokens"} {
+		req := httptest.NewRequest(http.MethodPost, path, strings.NewReader(`{"model":"grok","messages":[{"role":"user","content":"hi"}]}`))
+		req.Header.Set("Content-Type", "application/json")
+		w := httptest.NewRecorder()
 
-	router.ServeHTTP(w, req)
-	require.Equal(t, http.StatusNotFound, w.Code)
-	require.Contains(t, w.Body.String(), "Token counting is not supported for this platform")
+		countTokensRouter.ServeHTTP(w, req)
+		require.Equal(t, http.StatusOK, w.Code, "path=%s", path)
+		var response struct {
+			InputTokens int `json:"input_tokens"`
+		}
+		require.NoError(t, json.Unmarshal(w.Body.Bytes(), &response), "path=%s", path)
+		require.Positive(t, response.InputTokens, "path=%s", path)
+	}
 
 	for _, path := range []string{
 		"/v1/responses",