feat: support per-request model pricing

README.en.md

@@ -23,7 +23,7 @@ It relies on [CLIProxyAPI (CPA)](https://github.com/router-for-me/CLIProxyAPI) a
 - Analysis page for token trends, model/API Key/AI Provider composition, and hourly heatmaps
 - Standalone API Key usage page for querying usage by CPA API Key
 - Credentials page for Auth File and AI Provider usage, with credential quota lookup and refresh
-- Maintain model prices for cost estimation and reporting
+- Maintain model prices for cost estimation and reporting, including token pricing and per-request pricing
 - Optional password login protection, SQLite backups, Docker/Docker Compose, and systemd deployment
 
 ## Quick Start

README.md

@@ -23,7 +23,7 @@
 - 分析页面提供 Token 趋势、模型/API Key/AI Provider 构成和时段热力图
 - API Key 独立查询页，可按 CPA API Key 查看专属用量
 - 凭证页面展示 Auth File 与 AI Provider 使用情况，支持凭证限额查询与刷新
-- 可维护模型价格，用于成本估算和统计展示
+- 可维护模型价格，用于成本估算和统计展示，支持 Token 单价与按请求次数计费
 - 可选密码登录保护、SQLite 备份、Docker/Docker Compose 和 systemd 部署
 
 ## 快速开始

internal/api/pricing.go

@@ -18,6 +18,7 @@ type pricingEntryResponse struct {
 	PromptPricePer1M     float64 `json:"prompt_price_per_1m"`
 	CompletionPricePer1M float64 `json:"completion_price_per_1m"`
 	CachePricePer1M      float64 `json:"cache_price_per_1m"`
+	PricePerRequest      float64 `json:"price_per_request"`
 }
 
 type pricingListResponse struct {
@@ -29,6 +30,7 @@ type updatePricingRequest struct {
 	PromptPricePer1M     float64 `json:"prompt_price_per_1m"`
 	CompletionPricePer1M float64 `json:"completion_price_per_1m"`
 	CachePricePer1M      float64 `json:"cache_price_per_1m"`
+	PricePerRequest      float64 `json:"price_per_request"`
 }
 
 func registerPricingRoutes(router gin.IRoutes, pricingProvider service.PricingProvider) {
@@ -66,6 +68,7 @@ func registerPricingRoutes(router gin.IRoutes, pricingProvider service.PricingPr
 				PromptPricePer1M:     setting.PromptPricePer1M,
 				CompletionPricePer1M: setting.CompletionPricePer1M,
 				CachePricePer1M:      setting.CachePricePer1M,
+				PricePerRequest:      setting.PricePerRequest,
 			})
 		}
 		c.JSON(http.StatusOK, pricingListResponse{Pricing: response})
@@ -127,6 +130,7 @@ func updatePricing(c *gin.Context, pricingProvider service.PricingProvider, path
 		PromptPricePer1M:     request.PromptPricePer1M,
 		CompletionPricePer1M: request.CompletionPricePer1M,
 		CachePricePer1M:      request.CachePricePer1M,
+		PricePerRequest:      request.PricePerRequest,
 	})
 	if err != nil {
 		if strings.Contains(err.Error(), "has not been used") || strings.Contains(err.Error(), "required") || strings.Contains(err.Error(), "non-negative") {
@@ -142,5 +146,6 @@ func updatePricing(c *gin.Context, pricingProvider service.PricingProvider, path
 		PromptPricePer1M:     setting.PromptPricePer1M,
 		CompletionPricePer1M: setting.CompletionPricePer1M,
 		CachePricePer1M:      setting.CachePricePer1M,
+		PricePerRequest:      setting.PricePerRequest,
 	})
 }

internal/api/pricing_test.go

@@ -64,6 +64,7 @@ func TestPricingRoutesReturnConfiguredData(t *testing.T) {
 			PromptPricePer1M:     3,
 			CompletionPricePer1M: 15,
 			CachePricePer1M:      0.3,
+			PricePerRequest:      0.063,
 		}},
 	}, AuthConfig{}, nil, "")
 
@@ -77,7 +78,7 @@ func TestPricingRoutesReturnConfiguredData(t *testing.T) {
 	pricingReq := httptest.NewRequest(http.MethodGet, "/api/v1/pricing", nil)
 	pricingResp := httptest.NewRecorder()
 	router.ServeHTTP(pricingResp, pricingReq)
-	if pricingResp.Code != http.StatusOK || !contains(pricingResp.Body.String(), `"prompt_price_per_1m":3`) {
+	if pricingResp.Code != http.StatusOK || !contains(pricingResp.Body.String(), `"price_per_request":0.063`) {
 		t.Fatalf("unexpected pricing response: %d %s", pricingResp.Code, pricingResp.Body.String())
 	}
 }
@@ -89,16 +90,17 @@ func TestUpdatePricingRoute(t *testing.T) {
 			PromptPricePer1M:     3,
 			CompletionPricePer1M: 15,
 			CachePricePer1M:      0.3,
+			PricePerRequest:      0.063,
 		},
 	}
 	router := NewRouter(nil, nil, nil, provider, AuthConfig{}, nil, "")
 
-	req := httptest.NewRequest(http.MethodPut, "/api/v1/pricing/claude-sonnet", strings.NewReader(`{"prompt_price_per_1m":3,"completion_price_per_1m":15,"cache_price_per_1m":0.3}`))
+	req := httptest.NewRequest(http.MethodPut, "/api/v1/pricing/claude-sonnet", strings.NewReader(`{"prompt_price_per_1m":3,"completion_price_per_1m":15,"cache_price_per_1m":0.3,"price_per_request":0.063}`))
 	req.Header.Set("Content-Type", "application/json")
 	resp := httptest.NewRecorder()
 	router.ServeHTTP(resp, req)
 
-	if resp.Code != http.StatusOK || !contains(resp.Body.String(), `"model":"claude-sonnet"`) {
+	if resp.Code != http.StatusOK || !contains(resp.Body.String(), `"price_per_request":0.063`) {
 		t.Fatalf("unexpected update response: %d %s", resp.Code, resp.Body.String())
 	}
 }
@@ -114,15 +116,15 @@ func TestUpdatePricingRouteAcceptsModelInBody(t *testing.T) {
 	}
 	router := NewRouter(nil, nil, nil, provider, AuthConfig{}, nil, "")
 
-	req := httptest.NewRequest(http.MethodPut, "/api/v1/pricing", strings.NewReader(`{"model":"openai/gpt-4.1","prompt_price_per_1m":3,"completion_price_per_1m":15,"cache_price_per_1m":0.3}`))
+	req := httptest.NewRequest(http.MethodPut, "/api/v1/pricing", strings.NewReader(`{"model":"openai/gpt-4.1","prompt_price_per_1m":3,"completion_price_per_1m":15,"cache_price_per_1m":0.3,"price_per_request":0.063}`))
 	req.Header.Set("Content-Type", "application/json")
 	resp := httptest.NewRecorder()
 	router.ServeHTTP(resp, req)
 
 	if resp.Code != http.StatusOK || !contains(resp.Body.String(), `"model":"openai/gpt-4.1"`) {
 		t.Fatalf("unexpected update response: %d %s", resp.Code, resp.Body.String())
 	}
-	if provider.lastUpdate == nil || provider.lastUpdate.Model != "openai/gpt-4.1" {
+	if provider.lastUpdate == nil || provider.lastUpdate.Model != "openai/gpt-4.1" || provider.lastUpdate.PricePerRequest != 0.063 {
 		t.Fatalf("expected model from body to be passed through, got %+v", provider.lastUpdate)
 	}
 }

internal/entities/model_price_setting.go

@@ -9,6 +9,7 @@ type ModelPriceSetting struct {
 	PromptPricePer1M     float64
 	CompletionPricePer1M float64
 	CachePricePer1M      float64
+	PricePerRequest      float64
 	CreatedAt            time.Time `gorm:"serializer:storageTime"`
 	UpdatedAt            time.Time `gorm:"serializer:storageTime"`
 }

internal/helper/usage_cost.go

@@ -7,6 +7,7 @@ type UsageTokenCostInput struct {
 	InputTokens  int64
 	OutputTokens int64
 	CachedTokens int64
+	Requests     int64
 }
 
 // UsageEventRequiresPricing 判断事件是否包含需要价格表解释的计费 token。
@@ -29,6 +30,7 @@ func CalculateUsageEventCost(event entities.UsageEvent, pricing entities.ModelPr
 		InputTokens:  event.InputTokens,
 		OutputTokens: event.OutputTokens,
 		CachedTokens: event.CachedTokens,
+		Requests:     1,
 	}, pricing)
 }
 
@@ -46,11 +48,16 @@ func CalculateUsageTokenCost(input UsageTokenCostInput, pricing entities.ModelPr
 	if cachedTokens < 0 {
 		cachedTokens = 0
 	}
+	requests := input.Requests
+	if requests < 0 {
+		requests = 0
+	}
 	promptTokens := inputTokens - cachedTokens
 	if promptTokens < 0 {
 		promptTokens = 0
 	}
 	return (float64(promptTokens)/1_000_000.0)*pricing.PromptPricePer1M +
 		(float64(outputTokens)/1_000_000.0)*pricing.CompletionPricePer1M +
-		(float64(cachedTokens)/1_000_000.0)*pricing.CachePricePer1M
+		(float64(cachedTokens)/1_000_000.0)*pricing.CachePricePer1M +
+		float64(requests)*pricing.PricePerRequest
 }

internal/helper/usage_cost_test.go

@@ -31,3 +31,12 @@ func TestUsageEventRequiresPricingUsesBillableTokenFields(t *testing.T) {
 		t.Fatal("expected input tokens to require pricing")
 	}
 }
+
+func TestCalculateUsageTokenCostAddsPerRequestPricing(t *testing.T) {
+	pricing := entities.ModelPriceSetting{PricePerRequest: 0.063}
+	cost := CalculateUsageTokenCost(UsageTokenCostInput{Requests: 3}, pricing)
+	want := 0.189
+	if cost != want {
+		t.Fatalf("expected per-request cost %.3f, got %.3f", want, cost)
+	}
+}

internal/repository/db_test.go

@@ -54,8 +54,8 @@ func TestOpenDatabaseCreatesFreshDatabaseFromCurrentSchemaWithoutRunningMigratio
 	if err := db.Table("schema_migrations").Count(&count).Error; err != nil {
 		t.Fatalf("count schema migrations: %v", err)
 	}
-	if count != 29 {
-		t.Fatalf("expected fresh database to mark 29 migrations applied, got %d", count)
+	if count != 30 {
+		t.Fatalf("expected fresh database to mark 30 migrations applied, got %d", count)
 	}
 	if strings.Contains(logs.String(), "schema migration started") {
 		t.Fatalf("expected fresh database creation not to run version migrations, got logs:\n%s", logs.String())

internal/repository/dto/pricing.go

@@ -6,4 +6,5 @@ type ModelPriceSettingInput struct {
 	PromptPricePer1M     float64
 	CompletionPricePer1M float64
 	CachePricePer1M      float64
+	PricePerRequest      float64
 }

internal/repository/migration/20260528_model_price_per_request.go

@@ -0,0 +1,15 @@
+package migration
+
+import (
+	"fmt"
+
+	"cpa-usage-keeper/internal/entities"
+	"gorm.io/gorm"
+)
+
+func addModelPricePerRequestMigration(tx *gorm.DB) error {
+	if err := tx.AutoMigrate(&entities.ModelPriceSetting{}); err != nil {
+		return fmt.Errorf("auto migrate model price per request: %w", err)
+	}
+	return nil
+}

internal/repository/migration/migration.go

@@ -39,6 +39,7 @@ const (
 	migrationUsageOverviewRollupDimensions          = "20260518_usage_overview_rollup_dimensions"
 	migrationAddUsageEventReasoningEffort           = "20260519_add_usage_event_reasoning_effort"
 	migrationAddUsageEventQuotaWindowIndexes        = "20260525_add_usage_event_quota_window_indexes"
+	migrationAddModelPricePerRequest                = "20260528_add_model_price_per_request"
 )
 
 type schemaMigration struct {
@@ -122,6 +123,7 @@ func orderedMigrations() []databaseMigration {
 		{version: migrationUsageOverviewRollupDimensions, run: usageOverviewRollupDimensionsMigration, disableTransaction: true},
 		{version: migrationAddUsageEventReasoningEffort, run: addUsageEventReasoningEffortMigration},
 		{version: migrationAddUsageEventQuotaWindowIndexes, run: addUsageEventQuotaWindowIndexesMigration},
+		{version: migrationAddModelPricePerRequest, run: addModelPricePerRequestMigration},
 	}
 }
 

internal/repository/migration/migration_test.go

@@ -48,6 +48,7 @@ func TestOrderedMigrationsPreservesExecutionOrder(t *testing.T) {
 		"20260518_usage_overview_rollup_dimensions",
 		"20260519_add_usage_event_reasoning_effort",
 		"20260525_add_usage_event_quota_window_indexes",
+		"20260528_add_model_price_per_request",
 	}
 	if len(got) != len(want) {
 		t.Fatalf("expected ordered migrations %v, got %v", want, got)
@@ -112,6 +113,7 @@ func TestOpenDatabaseRunsSchemaMigrationsAndAddsUsageEventRedisFields(t *testing
 		"20260518_usage_overview_rollup_dimensions",
 		"20260519_add_usage_event_reasoning_effort",
 		"20260525_add_usage_event_quota_window_indexes",
+		"20260528_add_model_price_per_request",
 	}
 	if len(versions) != len(expected) {
 		t.Fatalf("expected migration versions %v, got %v", expected, versions)

internal/repository/pricing.go

@@ -47,7 +47,7 @@ func ListModelPriceSettings(db *gorm.DB) ([]entities.ModelPriceSetting, error) {
 	}
 
 	var settings []entities.ModelPriceSetting
-	if err := db.Select("ID", "Model", "PromptPricePer1M", "CompletionPricePer1M", "CachePricePer1M", "CreatedAt", "UpdatedAt").Order("model asc").Find(&settings).Error; err != nil {
+	if err := db.Select("ID", "Model", "PromptPricePer1M", "CompletionPricePer1M", "CachePricePer1M", "PricePerRequest", "CreatedAt", "UpdatedAt").Order("model asc").Find(&settings).Error; err != nil {
 		return nil, fmt.Errorf("list pricing settings: %w", err)
 	}
 	return settings, nil
@@ -64,7 +64,7 @@ func UpsertModelPriceSetting(db *gorm.DB, input dto.ModelPriceSettingInput) (*en
 	}
 
 	setting := &entities.ModelPriceSetting{}
-	if err := db.Select("ID", "Model", "PromptPricePer1M", "CompletionPricePer1M", "CachePricePer1M", "CreatedAt", "UpdatedAt").Where("model = ?", modelName).First(setting).Error; err != nil {
+	if err := db.Select("ID", "Model", "PromptPricePer1M", "CompletionPricePer1M", "CachePricePer1M", "PricePerRequest", "CreatedAt", "UpdatedAt").Where("model = ?", modelName).First(setting).Error; err != nil {
 		if err == gorm.ErrRecordNotFound {
 			setting = &entities.ModelPriceSetting{Model: modelName}
 		} else {
@@ -76,6 +76,7 @@ func UpsertModelPriceSetting(db *gorm.DB, input dto.ModelPriceSettingInput) (*en
 	setting.PromptPricePer1M = input.PromptPricePer1M
 	setting.CompletionPricePer1M = input.CompletionPricePer1M
 	setting.CachePricePer1M = input.CachePricePer1M
+	setting.PricePerRequest = input.PricePerRequest
 
 	if err := db.Save(setting).Error; err != nil {
 		return nil, fmt.Errorf("save pricing setting: %w", err)

internal/repository/query_projection_test.go

@@ -32,7 +32,7 @@ func TestRepositoryQueriesAvoidKnownFullEntityReads(t *testing.T) {
 		"Select(redisUsageInboxProcessingColumns).Where(\"status = ?\"",
 	)
 	assertFileContains(t, "pricing.go",
-		"Select(\"ID\", \"Model\", \"PromptPricePer1M\", \"CompletionPricePer1M\", \"CachePricePer1M\", \"CreatedAt\", \"UpdatedAt\")",
+		"Select(\"ID\", \"Model\", \"PromptPricePer1M\", \"CompletionPricePer1M\", \"CachePricePer1M\", \"PricePerRequest\", \"CreatedAt\", \"UpdatedAt\")",
 	)
 }
 

internal/repository/usage.go

@@ -1045,7 +1045,7 @@ func applyUsageOverviewHourlyStatToOverview(overview *dto.UsageOverviewRecord, r
 	// 小时 stats 是完整小时事实，可直接累计到 snapshot totals。
 	applyUsageOverviewHourlyStatToSnapshot(overview.Usage, row)
 	// cost 不入 stats 表，必须在读取时按当前价格表重新计算。
-	rowCost := helper.CalculateUsageTokenCost(helper.UsageTokenCostInput{InputTokens: row.InputTokens, OutputTokens: row.OutputTokens, CachedTokens: row.CachedTokens}, pricingByModel[strings.TrimSpace(row.Model)])
+	rowCost := helper.CalculateUsageTokenCost(helper.UsageTokenCostInput{InputTokens: row.InputTokens, OutputTokens: row.OutputTokens, CachedTokens: row.CachedTokens, Requests: row.RequestCount}, pricingByModel[strings.TrimSpace(row.Model)])
 	if _, ok := pricingByModel[strings.TrimSpace(row.Model)]; !ok && helper.UsageTokenInputRequiresPricing(helper.UsageTokenCostInput{InputTokens: row.InputTokens, OutputTokens: row.OutputTokens, CachedTokens: row.CachedTokens}) {
 		overview.Summary.CostAvailable = false
 	}
@@ -1069,7 +1069,7 @@ func applyUsageOverviewHourlyStatToHourlySeries(overview *dto.UsageOverviewRecor
 	if latestHourlyStart != nil && timeutil.NormalizeStorageTime(row.BucketStart).Before(*latestHourlyStart) {
 		return
 	}
-	rowCost := helper.CalculateUsageTokenCost(helper.UsageTokenCostInput{InputTokens: row.InputTokens, OutputTokens: row.OutputTokens, CachedTokens: row.CachedTokens}, pricingByModel[strings.TrimSpace(row.Model)])
+	rowCost := helper.CalculateUsageTokenCost(helper.UsageTokenCostInput{InputTokens: row.InputTokens, OutputTokens: row.OutputTokens, CachedTokens: row.CachedTokens, Requests: row.RequestCount}, pricingByModel[strings.TrimSpace(row.Model)])
 	hourKey, hourMinutes := usageOverviewBucket(timeutil.NormalizeStorageTime(row.BucketStart), false)
 	applyUsageOverviewStatToSeries(&overview.HourlySeries, row.Model, row.RequestCount, row.InputTokens, row.OutputTokens, row.CachedTokens, row.ReasoningTokens, row.TotalTokens, rowCost, hourKey, hourMinutes)
 }
@@ -1078,7 +1078,7 @@ func applyUsageOverviewHourlyStatToHourlySeries(overview *dto.UsageOverviewRecor
 func applyUsageOverviewDailyStatToOverview(overview *dto.UsageOverviewRecord, row entities.UsageOverviewDailyStat, bucketByDay bool, pricingByModel map[string]entities.ModelPriceSetting) {
 	// 天 stats 只覆盖完整本地天，不能用于非整天边界。
 	applyUsageOverviewDailyStatToSnapshot(overview.Usage, row)
-	rowCost := helper.CalculateUsageTokenCost(helper.UsageTokenCostInput{InputTokens: row.InputTokens, OutputTokens: row.OutputTokens, CachedTokens: row.CachedTokens}, pricingByModel[strings.TrimSpace(row.Model)])
+	rowCost := helper.CalculateUsageTokenCost(helper.UsageTokenCostInput{InputTokens: row.InputTokens, OutputTokens: row.OutputTokens, CachedTokens: row.CachedTokens, Requests: row.RequestCount}, pricingByModel[strings.TrimSpace(row.Model)])
 	if _, ok := pricingByModel[strings.TrimSpace(row.Model)]; !ok && helper.UsageTokenInputRequiresPricing(helper.UsageTokenCostInput{InputTokens: row.InputTokens, OutputTokens: row.OutputTokens, CachedTokens: row.CachedTokens}) {
 		overview.Summary.CostAvailable = false
 	}

internal/repository/usage_window_stats.go

@@ -26,6 +26,7 @@ type usageWindowTokenStats struct {
 	InputTokens  int64  `gorm:"column:input_tokens"`
 	OutputTokens int64  `gorm:"column:output_tokens"`
 	CachedTokens int64  `gorm:"column:cached_tokens"`
+	RequestCount int64  `gorm:"column:request_count"`
 }
 
 func SumUsageWindowStatsByAuthIndex(ctx context.Context, db *gorm.DB, authIndex string, start time.Time, end *time.Time) (UsageWindowStats, error) {
@@ -168,7 +169,7 @@ func sumRawUsageWindowTokenStats(db *gorm.DB, authIndex string, start time.Time,
 	// raw 查询只取 model 级汇总字段，避免把大量 usage_events 行读进 Go 内存。
 	query := db.Model(&entities.UsageEvent{}).
 		// SELECT 中只聚合 token/cost 需要的字段，不读取 raw_json 等大字段。
-		Select("model, COALESCE(SUM(total_tokens), 0) AS total_tokens, COALESCE(SUM(input_tokens), 0) AS input_tokens, COALESCE(SUM(output_tokens), 0) AS output_tokens, COALESCE(SUM(cached_tokens), 0) AS cached_tokens").
+		Select("model, COALESCE(SUM(total_tokens), 0) AS total_tokens, COALESCE(SUM(input_tokens), 0) AS input_tokens, COALESCE(SUM(output_tokens), 0) AS output_tokens, COALESCE(SUM(cached_tokens), 0) AS cached_tokens, COALESCE(COUNT(*), 0) AS request_count").
 		// auth_index 已经是唯一身份维度，这里不再额外按 auth_type 过滤。
 		Where("auth_index = ? AND timestamp >= ?", authIndex, timeutil.FormatStorageTime(start)).
 		// 按 model 分组，后续按 model 价格表计算 cost。
@@ -193,7 +194,7 @@ func sumHourlyUsageWindowTokenStats(db *gorm.DB, authIndex string, start time.Ti
 	// hourly 查询直接读取 overview 已经维护好的小时增量表。
 	query := db.Model(&entities.UsageOverviewHourlyStat{}).
 		// SELECT 中聚合 token/cost 需要的字段，保持和 raw 查询返回结构一致。
-		Select("model, COALESCE(SUM(total_tokens), 0) AS total_tokens, COALESCE(SUM(input_tokens), 0) AS input_tokens, COALESCE(SUM(output_tokens), 0) AS output_tokens, COALESCE(SUM(cached_tokens), 0) AS cached_tokens").
+		Select("model, COALESCE(SUM(total_tokens), 0) AS total_tokens, COALESCE(SUM(input_tokens), 0) AS input_tokens, COALESCE(SUM(output_tokens), 0) AS output_tokens, COALESCE(SUM(cached_tokens), 0) AS cached_tokens, COALESCE(SUM(request_count), 0) AS request_count").
 		// auth_index + bucket_start 范围可以使用现有 hourly auth_bucket 索引。
 		Where("auth_index = ? AND bucket_start >= ? AND bucket_start < ?", authIndex, timeutil.FormatStorageTime(start), timeutil.FormatStorageTime(end)).
 		// 按 model 分组，后续按 model 价格表计算 cost。
@@ -226,6 +227,7 @@ func mergeUsageWindowTokenStats(merged map[string]usageWindowTokenStats, rows []
 		current.OutputTokens += row.OutputTokens
 		// 累加 cached_tokens，用于 cache 成本计算并从 prompt 中扣除。
 		current.CachedTokens += row.CachedTokens
+		current.RequestCount += row.RequestCount
 		// 把合并后的 model 统计写回 map。
 		merged[model] = current
 	}
@@ -253,7 +255,7 @@ func usageWindowStatsFromTokenStats(rows []usageWindowTokenStats, pricingByModel
 		// model 名称按 trim 后查价格，保持和其它 Overview/Usage cost 逻辑一致。
 		pricing := pricingByModel[strings.TrimSpace(row.Model)]
 		// 使用统一 helper 按当前价格表计算该 model 的 cost。
-		stats.Cost += helper.CalculateUsageTokenCost(helper.UsageTokenCostInput{InputTokens: row.InputTokens, OutputTokens: row.OutputTokens, CachedTokens: row.CachedTokens}, pricing)
+		stats.Cost += helper.CalculateUsageTokenCost(helper.UsageTokenCostInput{InputTokens: row.InputTokens, OutputTokens: row.OutputTokens, CachedTokens: row.CachedTokens, Requests: row.RequestCount}, pricing)
 	}
 	// 返回最终窗口统计。
 	return stats