feat(models): update Cindy free catalog with native capabilities and standard pricing

backend/internal/service/account_test_service.go

@@ -37,6 +37,7 @@ import (
 	"github.com/gin-gonic/gin"
 	"github.com/google/uuid"
 	"github.com/tidwall/gjson"
+	"golang.org/x/sync/singleflight"
 )
 
 const (
@@ -152,24 +153,28 @@ func normalizeGrokAccountTestMode(mode string) string {
 
 // AccountTestService handles account testing operations
 type AccountTestService struct {
-	quotaActivity             *QuotaActivityService
-	accountRepo               AccountRepository
-	geminiTokenProvider       *GeminiTokenProvider
-	claudeTokenProvider       *ClaudeTokenProvider
-	grokTokenProvider         *GrokTokenProvider
-	antigravityGatewayService *AntigravityGatewayService
-	httpUpstream              HTTPUpstream
-	cfg                       *config.Config
-	settingService            *SettingService
-	tlsFPProfileService       *TLSFingerprintProfileService
-	openAIGatewayService      *OpenAIGatewayService
-	openaiGatewayService      *OpenAIGatewayService
-	modelMetadataRegistryMu   sync.Mutex
-	modelMetadataRegistry     map[string]modelsDevProvider
-	modelMetadataRegistryAt   time.Time
-	pluginManager             *PluginManager
-	agentIdentityTaskMu       sync.Mutex
-	agentIdentityWS           agentIdentityWSConnectionInvalidator
+	quotaActivity               *QuotaActivityService
+	accountRepo                 AccountRepository
+	geminiTokenProvider         *GeminiTokenProvider
+	claudeTokenProvider         *ClaudeTokenProvider
+	grokTokenProvider           *GrokTokenProvider
+	antigravityGatewayService   *AntigravityGatewayService
+	httpUpstream                HTTPUpstream
+	cfg                         *config.Config
+	settingService              *SettingService
+	tlsFPProfileService         *TLSFingerprintProfileService
+	openAIGatewayService        *OpenAIGatewayService
+	openaiGatewayService        *OpenAIGatewayService
+	modelMetadataRegistryMu     sync.Mutex
+	modelMetadataRegistry       map[string]modelsDevProvider
+	modelMetadataRegistryAt     time.Time
+	cindyMetadataRegistryMu     sync.Mutex
+	cindyMetadataRegistry       map[string]UpstreamModelMetadata
+	cindyMetadataRegistryAt     time.Time
+	cindyMetadataRegistryFlight singleflight.Group
+	pluginManager               *PluginManager
+	agentIdentityTaskMu         sync.Mutex
+	agentIdentityWS             agentIdentityWSConnectionInvalidator
 	// grokWSDialer is optional; realtime account tests use the default OpenAI-style
 	// WS dialer when nil (supports proxy + coder/websocket handshake).
 	grokWSDialer openAIWSClientDialer

backend/internal/service/billing_service.go

@@ -294,6 +294,7 @@ func NewBillingService(cfg *config.Config, pricingService *PricingService) *Bill
 // initFallbackPricing 初始化硬编码回退价格（当动态价格不可用时使用）
 // 价格单位：USD per token（与LiteLLM格式一致）
 func (s *BillingService) initFallbackPricing() {
+	addConversationStandardFallbackPricing(s.fallbackPrices)
 	// Claude 4.5 Opus
 	s.fallbackPrices["claude-opus-4.5"] = &ModelPricing{
 		InputPricePerToken:         5e-6,    // $5 per MTok
@@ -905,6 +906,12 @@ func (s *BillingService) initFallbackPricing() {
 // getFallbackPricing 根据模型系列获取回退价格
 func (s *BillingService) getFallbackPricing(model string) *ModelPricing {
 	modelLower := strings.ToLower(model)
+	switch modelLower {
+	case "hy3", "tencent/hy3", "hy4-preview", "tencent/hy4-preview",
+		"qwen3.8-27b", "qwen/qwen3.8-27b", "qwen3.8-flash", "qwen/qwen3.8-flash",
+		"qwen3.8-omni-flash", "qwen/qwen3.8-omni-flash", "muse-spark-1.3", "meta/muse-spark-1.3":
+		return s.fallbackPrices[modelLower]
+	}
 	if modelLower == "jev-latest" {
 		return s.fallbackPrices["jev-latest"]
 	}
@@ -1262,7 +1269,7 @@ func (s *BillingService) GetModelPricingWithChannel(model string, channelPricing
 	if err != nil {
 		return nil, err
 	}
-	if channelPricing == nil {
+	if channelPricing == nil || channelPricing.IsTokenAllowlistOnly() {
 		return pricing, nil
 	}
 	// 防止修改 fallbackPrices 中的共享指针

backend/internal/service/channel.go

@@ -110,6 +110,16 @@ type ChannelModelPricing struct {
 	UpdatedAt                  time.Time           `json:"updated_at,omitempty"`
 }
 
+// IsTokenAllowlistOnly distinguishes a model restriction from a price override.
+// An explicitly configured zero remains an override.
+func (p *ChannelModelPricing) IsTokenAllowlistOnly() bool {
+	return p != nil && (p.BillingMode == "" || p.BillingMode == BillingModeToken) &&
+		p.InputPrice == nil && p.OutputPrice == nil && p.CacheWritePrice == nil && p.CacheWrite1hPrice == nil &&
+		p.CacheReadPrice == nil && p.FastMultiplier == nil && p.FlexMultiplier == nil &&
+		len(p.ReasoningEffortMultipliers) == 0 && p.ImageInputPrice == nil && p.ImageOutputPrice == nil &&
+		p.PerRequestPrice == nil && len(p.Intervals) == 0 && p.TimePricing == nil
+}
+
 // ChannelTimePricing 渠道模型定价的分时倍率配置。
 type ChannelTimePricing struct {
 	Timezone     string                     `json:"timezone"`

backend/internal/service/conversation_standard_pricing.go

@@ -0,0 +1,31 @@
+package service
+
+// Exact international USD API prices verified 2026-10-04. These enter the
+// ordinary fallback table; a dynamic catalog price remains authoritative.
+// Tencent: https://intl.cloud.tencent.com/zh/document/product/1300/78937
+// Qwen: https://www.alibabacloud.com/help/en/model-studio/qwen3-8-27b
+//
+//	https://www.alibabacloud.com/help/en/model-studio/qwen3-8-flash
+//	https://www.alibabacloud.com/help/en/model-studio/model-pricing
+//
+// Meta: https://dev.meta.ai/docs/pricing-rate-limits
+func addConversationStandardFallbackPricing(prices map[string]*ModelPricing) {
+	for _, row := range []struct {
+		name, provider                       string
+		input, output, cacheRead, cacheWrite float64 // USD / million tokens
+	}{
+		{"hy3", "tencent", 0.132, 0.528, 0.033, 0},
+		{"hy4-preview", "tencent", 0.834, 2.501, 0.042, 0},
+		// The single cached-input field uses the implicit-cache rate. Explicit
+		// cache creation has its own published rate; no TTL ladder is assumed.
+		{"qwen3.8-27b", "qwen", 0.50, 3.00, 0.10, 0.625},
+		{"qwen3.8-flash", "qwen", 0.15, 0.47, 0.016, 0.20},
+		{"qwen3.8-omni-flash", "qwen", 0.15, 0.47, 0.016, 0},
+		{"muse-spark-1.3", "meta", 1.25, 4.25, 0.15, 0},
+	} {
+		price := &ModelPricing{InputPricePerToken: row.input / 1e6, OutputPricePerToken: row.output / 1e6,
+			CacheReadPricePerToken: row.cacheRead / 1e6, CacheCreationPricePerToken: row.cacheWrite / 1e6}
+		prices[row.name] = price
+		prices[row.provider+"/"+row.name] = price
+	}
+}

backend/internal/service/conversation_standard_pricing_test.go

@@ -0,0 +1,65 @@
+//go:build unit
+
+package service
+
+import (
+	"context"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestConversationStandardPricingExactFallbackAndDynamicPriority(t *testing.T) {
+	bs := &BillingService{fallbackPrices: make(map[string]*ModelPricing)}
+	bs.initFallbackPricing()
+	for _, row := range []struct {
+		model, namespace           string
+		input, output, read, write float64
+	}{
+		{"hy3", "tencent", .132, .528, .033, 0}, {"hy4-preview", "tencent", .834, 2.501, .042, 0},
+		{"qwen3.8-27b", "qwen", .5, 3, .1, .625}, {"qwen3.8-flash", "qwen", .15, .47, .016, .2},
+		{"qwen3.8-omni-flash", "qwen", .15, .47, .016, 0}, {"muse-spark-1.3", "meta", 1.25, 4.25, .15, 0},
+	} {
+		for _, model := range []string{row.model, row.namespace + "/" + row.model} {
+			price, err := bs.GetModelPricing(model)
+			require.NoError(t, err, model)
+			require.InDelta(t, row.input/1e6, price.InputPricePerToken, 1e-14)
+			require.InDelta(t, row.output/1e6, price.OutputPricePerToken, 1e-14)
+			require.InDelta(t, row.read/1e6, price.CacheReadPricePerToken, 1e-14)
+			require.InDelta(t, row.write/1e6, price.CacheCreationPricePerToken, 1e-14)
+		}
+	}
+	for _, unknown := range []string{"qwen3.8-unknown", "qwen3.8-flash-next", "hy4-preview-unknown", "muse-spark-1.3-contributor", "other/hy3"} {
+		_, err := bs.GetModelPricing(unknown)
+		require.ErrorIs(t, err, ErrModelPricingUnavailable, unknown)
+	}
+	bs.pricingService = newStubPricingServiceFromJSON(t, `{"hy3":{"input_cost_per_token":0.000001,"output_cost_per_token":0.000002,"litellm_provider":"tencent","mode":"chat"}}`)
+	price, err := bs.GetModelPricing("hy3")
+	require.NoError(t, err)
+	require.Equal(t, 1e-6, price.InputPricePerToken)
+}
+
+func TestTokenAllowlistOnlyPreservesDefaultPricing(t *testing.T) {
+	bs := newTestBillingServiceForResolver()
+	base := bs.fallbackPrices["claude-sonnet-4"]
+	fast, flex := 2.0, .5
+	base.FastMultiplier, base.FlexMultiplier = &fast, &flex
+	base.ReasoningEffortMultipliers = map[string]float64{"high": 1.5}
+	card := ChannelModelPricing{Platform: PlatformOpenAI, Models: []string{"claude-sonnet-4"}, BillingMode: BillingModeToken}
+	require.True(t, card.IsTokenAllowlistOnly())
+	cs := newChannelServiceWithPricings(12, []ChannelModelPricing{card})
+	groupID := int64(12)
+	resolved := NewModelPricingResolver(cs, bs).Resolve(context.Background(), PricingInput{Model: "claude-sonnet-4", GroupID: &groupID})
+	require.Equal(t, PricingSourceLiteLLM, resolved.Source)
+	require.Equal(t, base, resolved.BasePricing)
+	price, err := bs.GetModelPricingWithChannel("claude-sonnet-4", &card)
+	require.NoError(t, err)
+	require.Equal(t, base, price)
+	zero := 0.0
+	card.InputPrice = &zero
+	require.False(t, card.IsTokenAllowlistOnly())
+	price, err = bs.GetModelPricingWithChannel("claude-sonnet-4", &card)
+	require.NoError(t, err)
+	require.Zero(t, price.InputPricePerToken)
+	require.Equal(t, 3e-6, base.InputPricePerToken, "explicit overrides do not mutate shared defaults")
+}

backend/internal/service/model_pricing_resolver.go

@@ -87,6 +87,9 @@ func (r *ModelPricingResolver) Resolve(ctx context.Context, input PricingInput)
 	var chPricing *ChannelModelPricing
 	if input.GroupID != nil && r.channelService != nil {
 		chPricing = r.lookupChannelPricingNormalized(ctx, *input.GroupID, input.Model)
+		if chPricing.IsTokenAllowlistOnly() {
+			chPricing = nil
+		}
 		if chPricing != nil {
 			mode := chPricing.BillingMode
 			if mode == "" {
@@ -225,7 +228,7 @@ func (r *ModelPricingResolver) lookupChannelPricingNormalized(ctx context.Contex
 // applyChannelOverrides 应用渠道定价覆盖
 func (r *ModelPricingResolver) applyChannelOverrides(ctx context.Context, groupID int64, model string, resolved *ResolvedPricing) {
 	chPricing := r.lookupChannelPricingNormalized(ctx, groupID, model)
-	if chPricing == nil {
+	if chPricing == nil || chPricing.IsTokenAllowlistOnly() {
 		return
 	}
 
@@ -246,6 +249,9 @@ func (r *ModelPricingResolver) applyChannelOverrides(ctx context.Context, groupI
 
 // applyTokenOverrides 应用 token 模式的渠道覆盖
 func (r *ModelPricingResolver) applyTokenOverrides(chPricing *ChannelModelPricing, resolved *ResolvedPricing) {
+	if chPricing.IsTokenAllowlistOnly() {
+		return
+	}
 	if resolved.BasePricing == nil {
 		resolved.BasePricing = &ModelPricing{}
 	} else {