Commit eb99ab1b by CaIon

feat(billing): add built-in expression pricing for gpt-6-astra

parent d5803532
...@@ -3,6 +3,7 @@ package controller ...@@ -3,6 +3,7 @@ package controller
import ( import (
"fmt" "fmt"
"net/http" "net/http"
"slices"
"sort" "sort"
"strconv" "strconv"
"strings" "strings"
...@@ -85,7 +86,7 @@ func GetOptions(c *gin.Context) { ...@@ -85,7 +86,7 @@ func GetOptions(c *gin.Context) {
optionValues := make(map[string]string) optionValues := make(map[string]string)
common.OptionMapRWMutex.Lock() common.OptionMapRWMutex.Lock()
for k, v := range common.OptionMap { for k, v := range common.OptionMap {
if k == "theme.frontend" { if k == "theme.frontend" || k == "billing_setting.billing_mode" || k == "billing_setting.billing_expr" {
continue continue
} }
value := common.Interface2String(v) value := common.Interface2String(v)
...@@ -101,14 +102,24 @@ func GetOptions(c *gin.Context) { ...@@ -101,14 +102,24 @@ func GetOptions(c *gin.Context) {
Key: k, Key: k,
Value: value, Value: value,
}) })
for _, optionKey := range completionRatioMetaOptionKeys { if slices.Contains(completionRatioMetaOptionKeys, k) {
if optionKey == k { optionValues[k] = value
optionValues[k] = value
break
}
} }
} }
common.OptionMapRWMutex.Unlock() common.OptionMapRWMutex.Unlock()
// Display the same effective expressions used by pricing and settlement,
// including built-in defaults absent from persisted administrator options.
for key, values := range map[string]map[string]string{
"billing_setting.billing_mode": billing_setting.GetBillingModeCopy(),
"billing_setting.billing_expr": billing_setting.GetBillingExprCopy(),
} {
encoded, err := common.Marshal(values)
if err != nil {
c.JSON(http.StatusInternalServerError, gin.H{"success": false, "message": err.Error()})
return
}
options = append(options, &model.Option{Key: key, Value: string(encoded)})
}
options = append(options, &model.Option{ options = append(options, &model.Option{
Key: "CompletionRatioMeta", Key: "CompletionRatioMeta",
Value: buildCompletionRatioMetaValue(optionValues), Value: buildCompletionRatioMetaValue(optionValues),
......
package openai package openai
var ModelList = []string{ var ModelList = []string{
"gpt-6-astra",
"gpt-3.5-turbo", "gpt-3.5-turbo-0613", "gpt-3.5-turbo-1106", "gpt-3.5-turbo-0125", "gpt-3.5-turbo", "gpt-3.5-turbo-0613", "gpt-3.5-turbo-1106", "gpt-3.5-turbo-0125",
"gpt-3.5-turbo-16k", "gpt-3.5-turbo-16k-0613", "gpt-3.5-turbo-16k", "gpt-3.5-turbo-16k-0613",
"gpt-3.5-turbo-instruct", "gpt-3.5-turbo-instruct-0914", "gpt-3.5-turbo-instruct", "gpt-3.5-turbo-instruct-0914",
......
package billing_setting
// Built-in token prices use actual USD per million tokens. Keep new model
// defaults here instead of splitting them across the legacy ratio tables.
var builtinBillingExpr = map[string]string{
// https://developers.openai.com/api/docs/models/gpt-6-astra
// Standard pricing; the long-context rates apply to the whole request.
// Do not infer service-tier discounts from incoming request parameters:
// channels filter service_tier by default, so it may not reach the upstream.
"gpt-6-astra": `len <= 272000 ? tier("standard", p * 10 + c * 50 + cr * 1 + cc * 12.5) : tier("long_context", p * 20 + c * 75 + cr * 2 + cc * 25)`,
}
package billing_setting_test
import (
"net/http/httptest"
"testing"
"github.com/QuantumNous/new-api/common"
"github.com/QuantumNous/new-api/controller"
"github.com/QuantumNous/new-api/model"
"github.com/QuantumNous/new-api/pkg/billingexpr"
"github.com/QuantumNous/new-api/relaykit/dto"
"github.com/QuantumNous/new-api/service"
"github.com/QuantumNous/new-api/setting/billing_setting"
"github.com/QuantumNous/new-api/setting/config"
"github.com/QuantumNous/new-api/setting/ratio_setting"
"github.com/gin-gonic/gin"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
)
func TestGPT6AstraBuiltinBilling(t *testing.T) {
settings := config.GlobalConfig.Get("billing_setting").(*billing_setting.BillingSetting)
saved := *settings
savedRatios, savedPrices := ratio_setting.ModelRatio2JSONString(), ratio_setting.ModelPrice2JSONString()
savedOptions := common.OptionMap
t.Cleanup(func() {
*settings, common.OptionMap = saved, savedOptions
require.NoError(t, ratio_setting.UpdateModelRatioByJSONString(savedRatios))
require.NoError(t, ratio_setting.UpdateModelPriceByJSONString(savedPrices))
})
common.OptionMap = map[string]string{"billing_setting.billing_mode": `{}`, "billing_setting.billing_expr": `{}`}
require.NoError(t, config.GlobalConfig.LoadFromDB(common.OptionMap))
require.NoError(t, ratio_setting.UpdateModelRatioByJSONString(`{}`))
require.NoError(t, ratio_setting.UpdateModelPriceByJSONString(`{}`))
assert.Equal(t, billing_setting.BillingModeTieredExpr, billing_setting.GetBillingMode("gpt-6-astra"))
expression, ok := billing_setting.GetBillingExpr("gpt-6-astra")
require.True(t, ok)
for _, tc := range []struct {
name string
input, output, cached, written int
request string
quota int
}{
{"standard", 1000, 100, 0, 0, `{}`, 7500},
{"client flex cannot discount standard pricing", 1000, 100, 0, 0, `{"service_tier":"flex"}`, 7500},
{"cache at context boundary", 272000, 1000, 200000, 20000, `{}`, 510000},
{"whole request above boundary", 272001, 1000, 200000, 20000, `{}`, 1007510},
} {
t.Run(tc.name, func(t *testing.T) {
usage := &dto.Usage{
PromptTokens: tc.input, CompletionTokens: tc.output,
PromptTokensDetails: dto.InputTokenDetails{CachedTokens: tc.cached, CacheWriteTokens: tc.written},
}
params := service.BuildTieredTokenParams(usage, false, billingexpr.UsedVars(expression))
result, err := billingexpr.ComputeTieredQuotaWithRequest(&billingexpr.BillingSnapshot{
ExprString: expression, GroupRatio: 1, QuotaPerUnit: 500000,
}, params, billingexpr.RequestInput{Body: []byte(tc.request)})
require.NoError(t, err)
assert.Equal(t, tc.quota, result.ActualQuotaAfterGroup)
})
}
t.Run("admin options expose defaults without persisting them", func(t *testing.T) {
recorder := httptest.NewRecorder()
ctx, _ := gin.CreateTestContext(recorder)
controller.GetOptions(ctx)
var response struct {
Success bool
Data []model.Option
}
require.NoError(t, common.Unmarshal(recorder.Body.Bytes(), &response))
require.True(t, response.Success)
found := map[string]string{}
for _, option := range response.Data {
if _, ok := common.OptionMap[option.Key]; ok {
assert.NotContains(t, found, option.Key)
var values map[string]string
require.NoError(t, common.UnmarshalJsonStr(option.Value, &values))
found[option.Key] = values["gpt-6-astra"]
assert.Equal(t, `{}`, common.OptionMap[option.Key])
}
}
assert.Equal(t, map[string]string{"billing_setting.billing_mode": "tiered_expr", "billing_setting.billing_expr": expression}, found)
})
for _, tc := range []struct {
name, mode, expr, ratios, prices, wantMode string
}{
{"custom expression overrides legacy price", "tiered_expr", "p * 7", `{"gpt-6-astra":8}`, `{}`, "tiered_expr"},
{"explicit ratio mode", "ratio", "", `{}`, `{}`, "ratio"},
{"existing free token price", "", "", `{"gpt-6-astra":0}`, `{}`, "ratio"},
{"existing per-call price", "", "", `{}`, `{"gpt-6-astra":0.1}`, "ratio"},
} {
t.Run(tc.name, func(t *testing.T) {
*settings = billing_setting.BillingSetting{BillingMode: map[string]string{}, BillingExpr: map[string]string{}}
if tc.mode != "" {
settings.BillingMode["gpt-6-astra"] = tc.mode
}
if tc.expr != "" {
settings.BillingExpr["gpt-6-astra"] = tc.expr
}
require.NoError(t, ratio_setting.UpdateModelRatioByJSONString(tc.ratios))
require.NoError(t, ratio_setting.UpdateModelPriceByJSONString(tc.prices))
assert.Equal(t, tc.wantMode, billing_setting.GetBillingMode("gpt-6-astra"))
actual, ok := billing_setting.GetBillingExpr("gpt-6-astra")
assert.Equal(t, tc.expr, actual)
assert.Equal(t, tc.expr != "", ok)
})
}
}
...@@ -6,11 +6,12 @@ import ( ...@@ -6,11 +6,12 @@ import (
"sort" "sort"
"github.com/QuantumNous/new-api/common" "github.com/QuantumNous/new-api/common"
"github.com/QuantumNous/new-api/relaykit/dto"
"github.com/QuantumNous/new-api/pkg/billingexpr" "github.com/QuantumNous/new-api/pkg/billingexpr"
"github.com/QuantumNous/new-api/pkg/jsplugin" "github.com/QuantumNous/new-api/pkg/jsplugin"
relaycommon "github.com/QuantumNous/new-api/relay/common" relaycommon "github.com/QuantumNous/new-api/relay/common"
"github.com/QuantumNous/new-api/relaykit/dto"
"github.com/QuantumNous/new-api/setting/config" "github.com/QuantumNous/new-api/setting/config"
"github.com/QuantumNous/new-api/setting/ratio_setting"
"github.com/samber/lo" "github.com/samber/lo"
) )
...@@ -46,20 +47,52 @@ func GetBillingMode(model string) string { ...@@ -46,20 +47,52 @@ func GetBillingMode(model string) string {
if mode, ok := billingSetting.BillingMode[model]; ok { if mode, ok := billingSetting.BillingMode[model]; ok {
return mode return mode
} }
if _, ok := builtinBillingExpr[model]; ok {
// Existing administrator-configured legacy prices take precedence over
// a newly introduced built-in expression unless a mode was explicit.
if ratio_setting.HasConfiguredModelRatio(model) {
return BillingModeRatio
}
if _, configured := ratio_setting.GetModelPrice(model, false); configured {
return BillingModeRatio
}
return BillingModeTieredExpr
}
return BillingModeRatio return BillingModeRatio
} }
func GetBillingExpr(model string) (string, bool) { func GetBillingExpr(model string) (string, bool) {
expr, ok := billingSetting.BillingExpr[model] if expr, ok := billingSetting.BillingExpr[model]; ok {
return expr, ok return expr, true
}
if GetBillingMode(model) == BillingModeTieredExpr {
expr, ok := builtinBillingExpr[model]
return expr, ok
}
return "", false
} }
func GetBillingModeCopy() map[string]string { func GetBillingModeCopy() map[string]string {
return lo.Assign(billingSetting.BillingMode) modes := lo.Assign(billingSetting.BillingMode)
for model := range builtinBillingExpr {
if _, configured := modes[model]; !configured && GetBillingMode(model) == BillingModeTieredExpr {
modes[model] = BillingModeTieredExpr
}
}
return modes
} }
func GetBillingExprCopy() map[string]string { func GetBillingExprCopy() map[string]string {
return lo.Assign(billingSetting.BillingExpr) expressions := lo.Assign(billingSetting.BillingExpr)
for model := range builtinBillingExpr {
if _, configured := expressions[model]; configured {
continue
}
if expression, ok := GetBillingExpr(model); ok {
expressions[model] = expression
}
}
return expressions
} }
func GetPricingSyncData(base map[string]any) map[string]any { func GetPricingSyncData(base map[string]any) map[string]any {
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or sign in to comment