feat: sync deployed v1.0.0-rc.38-custom.42
Sync production source including routing, billing, Astra detection and public uptime. Update custom build instructions and license packaging.
This commit is contained in:
cneicy committed
1 parent
aa6c3f50f4
commit
ceddfd3d6b
828 files changed
+122253
-13702
No files matched your search
@@ -1,22 +1,138 @@
|
||||
package relay
|
||||
|
||||
import (
|
||||
"errors"
|
||||
"io"
|
||||
"net/http"
|
||||
"net/http/httptest"
|
||||
"strings"
|
||||
"testing"
|
||||
|
||||
"github.com/QuantumNous/new-api/common"
|
||||
"github.com/QuantumNous/new-api/constant"
|
||||
"github.com/QuantumNous/new-api/pkg/billingexpr"
|
||||
relaycommon "github.com/QuantumNous/new-api/relay/common"
|
||||
relayconstant "github.com/QuantumNous/new-api/relay/constant"
|
||||
"github.com/QuantumNous/new-api/relay/helper"
|
||||
"github.com/QuantumNous/new-api/relaykit/dto"
|
||||
kitreasoning "github.com/QuantumNous/new-api/relaykit/relayconvert/reasoning"
|
||||
"github.com/QuantumNous/new-api/relaykit/types"
|
||||
"github.com/QuantumNous/new-api/service"
|
||||
"github.com/QuantumNous/new-api/setting/model_setting"
|
||||
hosttypes "github.com/QuantumNous/new-api/types"
|
||||
"github.com/gin-gonic/gin"
|
||||
"github.com/stretchr/testify/assert"
|
||||
"github.com/stretchr/testify/require"
|
||||
"github.com/tidwall/gjson"
|
||||
)
|
||||
|
||||
type imageReservation struct {
|
||||
held, limit int
|
||||
}
|
||||
|
||||
func (s *imageReservation) Reserve(quota int) error {
|
||||
if quota > s.limit {
|
||||
return errors.New("insufficient image quota")
|
||||
}
|
||||
s.held = max(s.held, quota)
|
||||
return nil
|
||||
}
|
||||
func (s *imageReservation) GetPreConsumedQuota() int { return s.held }
|
||||
func (*imageReservation) Settle(int) error { return nil }
|
||||
func (*imageReservation) Refund(*gin.Context) {}
|
||||
func (*imageReservation) NeedsRefund() bool { return false }
|
||||
|
||||
func TestImageRequestReservesFinalQuantityBeforeUpstream(t *testing.T) {
|
||||
service.InitHttpClient()
|
||||
for _, tc := range []struct {
|
||||
name, body string
|
||||
tiered, passThrough, insufficient, retryToOpenAI bool
|
||||
override any
|
||||
count, status int
|
||||
}{
|
||||
{name: "Ali nested count", body: `{"model":"z-image","n":1,"parameters":{"n":4}}`, count: 4},
|
||||
{name: "Ali empty parameters", body: `{"model":"z-image","n":2,"parameters":{}}`, count: 2},
|
||||
{name: "legacy channel quantity override", body: `{"model":"z-image","n":1}`, override: 4, count: 4},
|
||||
{name: "expression channel quantity override", body: `{"model":"z-image","n":1}`, override: 4, count: 4, tiered: true},
|
||||
{name: "pass-through quantity", body: `{"model":"z-image","n":2,"parameters":{}}`, count: 2, passThrough: true},
|
||||
{name: "zero override rejected", body: `{"model":"z-image","n":1}`, override: 0, status: http.StatusBadRequest},
|
||||
{name: "oversized override rejected", body: `{"model":"z-image","n":1}`, override: 129, status: http.StatusBadRequest},
|
||||
{name: "insufficient reservation blocks upstream", body: `{"model":"z-image","n":1}`, override: 4, count: 4, insufficient: true, status: http.StatusForbidden},
|
||||
{name: "retry drops Ali quantity and surcharge", body: `{"model":"z-image","n":1,"parameters":{"n":4,"prompt_extend":true}}`, count: 1, retryToOpenAI: true},
|
||||
} {
|
||||
t.Run(tc.name, func(t *testing.T) {
|
||||
received := make(chan []byte, 1)
|
||||
upstream := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
|
||||
body, err := io.ReadAll(r.Body)
|
||||
if err == nil {
|
||||
received <- body
|
||||
}
|
||||
w.Header().Set("Content-Type", "application/json")
|
||||
w.WriteHeader(http.StatusBadGateway)
|
||||
_, _ = io.WriteString(w, `{"error":{"message":"fixture upstream failure","type":"upstream_error"}}`)
|
||||
}))
|
||||
t.Cleanup(upstream.Close)
|
||||
c, _ := gin.CreateTestContext(httptest.NewRecorder())
|
||||
c.Request = httptest.NewRequest(http.MethodPost, "/v1/images/generations", strings.NewReader(tc.body))
|
||||
c.Request.Header.Set("Content-Type", "application/json")
|
||||
channel := constant.ChannelTypeAli
|
||||
if tc.retryToOpenAI {
|
||||
channel = constant.ChannelTypeOpenAI
|
||||
}
|
||||
common.SetContextKey(c, constant.ContextKeyChannelType, channel)
|
||||
common.SetContextKey(c, constant.ContextKeyChannelBaseUrl, upstream.URL)
|
||||
common.SetContextKey(c, constant.ContextKeyChannelSetting, dto.ChannelSettings{PassThroughBodyEnabled: tc.passThrough})
|
||||
if tc.override != nil {
|
||||
common.SetContextKey(c, constant.ContextKeyChannelParamOverride, map[string]any{"operations": []any{map[string]any{"path": "parameters.n", "mode": "set", "value": tc.override}}})
|
||||
}
|
||||
request, err := helper.GetAndValidOpenAIImageRequest(c, relayconstant.RelayModeImagesGenerations)
|
||||
require.NoError(t, err)
|
||||
reservation := &imageReservation{held: 20000, limit: 500000}
|
||||
if tc.insufficient {
|
||||
reservation.limit = reservation.held
|
||||
}
|
||||
info := &relaycommon.RelayInfo{Request: request, OriginModelName: "z-image", RelayMode: relayconstant.RelayModeImagesGenerations,
|
||||
RequestURLPath: c.Request.URL.Path, Billing: reservation,
|
||||
PriceData: hosttypes.PriceData{UsePrice: true, ModelPrice: 0.04, GroupRatioInfo: hosttypes.GroupRatioInfo{GroupRatio: 1}},
|
||||
}
|
||||
if tc.tiered {
|
||||
expr := `tier("image", fixed(0.04)) * image_count`
|
||||
info.TieredBillingSnapshot = &billingexpr.BillingSnapshot{BillingMode: "tiered_expr", ExprString: expr, ExprHash: billingexpr.ExprHashString(expr), GroupRatio: 1, QuotaPerUnit: common.QuotaPerUnit, EstimatedImageCount: common.GetPointer(1)}
|
||||
info.BillingRequestInput = &billingexpr.RequestInput{Body: []byte(tc.body), ImageCount: common.GetPointer(1)}
|
||||
info.PriceData.UsePrice = false
|
||||
}
|
||||
if tc.retryToOpenAI {
|
||||
reservation.held = 160000
|
||||
info.PriceData.AddOtherRatio("n", 4)
|
||||
info.PriceData.AddOtherRatio("prompt_extend", 2)
|
||||
info.BillingImageCount = common.GetPointer(3)
|
||||
}
|
||||
apiErr := ImageHelper(c, info)
|
||||
require.NotNil(t, apiErr)
|
||||
if tc.status != 0 {
|
||||
assert.Equal(t, tc.status, apiErr.StatusCode)
|
||||
assert.Empty(t, received, "no upstream submission before quantity validation and reservation")
|
||||
return
|
||||
}
|
||||
assert.Equal(t, http.StatusBadGateway, apiErr.StatusCode)
|
||||
require.Len(t, received, 1)
|
||||
body := <-received
|
||||
path := "parameters.n"
|
||||
if tc.retryToOpenAI {
|
||||
path = "n"
|
||||
}
|
||||
assert.Equal(t, int64(tc.count), gjson.GetBytes(body, path).Int())
|
||||
assert.Equal(t, tc.count*20000, info.PriceData.QuotaToPreConsume)
|
||||
assert.GreaterOrEqual(t, reservation.held, info.PriceData.QuotaToPreConsume)
|
||||
assert.Nil(t, info.BillingImageCount)
|
||||
if tc.tiered {
|
||||
assert.Equal(t, tc.body, string(info.BillingRequestInput.Body))
|
||||
assert.Equal(t, 1, *info.BillingRequestInput.ImageCount)
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
func TestOptInSafeToolLossRejectedAsBadRequestWithAdminDiagnostics(t *testing.T) {
|
||||
gin.SetMode(gin.TestMode)
|
||||
c, _ := gin.CreateTestContext(httptest.NewRecorder())
|
||||
@@ -96,3 +212,90 @@ func hasHostDiagnosticCode(diagnostics []types.ConversionDiagnostic, code string
|
||||
}
|
||||
return false
|
||||
}
|
||||
|
||||
func TestGeminiThinkingControlsConvertBestEffort(t *testing.T) {
|
||||
gin.SetMode(gin.TestMode)
|
||||
settings := model_setting.GetGeminiSettings()
|
||||
originalAdapter := settings.ThinkingAdapterEnabled
|
||||
t.Cleanup(func() { settings.ThinkingAdapterEnabled = originalAdapter })
|
||||
|
||||
t.Run("openai thinking_budget on gemini 3 becomes thinkingLevel", func(t *testing.T) {
|
||||
settings.ThinkingAdapterEnabled = false
|
||||
c, _ := gin.CreateTestContext(httptest.NewRecorder())
|
||||
c.Request = httptest.NewRequest(http.MethodPost, "/v1/chat/completions", nil)
|
||||
req := &dto.GeneralOpenAIRequest{
|
||||
Model: "gemini-3.1-pro-preview-thinking",
|
||||
Messages: []dto.Message{{Role: "user", Content: "hello"}},
|
||||
ExtraBody: []byte(`{"google":{"thinking_config":{"thinking_budget":8192}}}`),
|
||||
}
|
||||
info := &relaycommon.RelayInfo{
|
||||
RelayFormat: types.RelayFormatOpenAI,
|
||||
OriginModelName: "gemini-3.1-pro-preview-thinking",
|
||||
Request: req,
|
||||
ChannelMeta: &relaycommon.ChannelMeta{UpstreamModelName: "gemini-3.1-pro-preview"},
|
||||
}
|
||||
require.NoError(t, helper.ApplyReasoningModelSuffix(c, info, req))
|
||||
|
||||
result, err := service.ConvertRequest(c, info, types.RelayFormatGemini, req)
|
||||
require.NoError(t, err)
|
||||
converted, ok := result.Value.(*dto.GeminiChatRequest)
|
||||
require.True(t, ok)
|
||||
require.NotNil(t, converted.GenerationConfig.ThinkingConfig)
|
||||
assert.Equal(t, "medium", converted.GenerationConfig.ThinkingConfig.ThinkingLevel)
|
||||
assert.Nil(t, converted.GenerationConfig.ThinkingConfig.ThinkingBudget)
|
||||
assert.True(t, hasHostDiagnosticCode(info.ConversionDiagnostics(), "gemini_budget_to_level"))
|
||||
})
|
||||
|
||||
t.Run("thinking alias with native thinking_level keeps the level", func(t *testing.T) {
|
||||
settings.ThinkingAdapterEnabled = true
|
||||
c, _ := gin.CreateTestContext(httptest.NewRecorder())
|
||||
c.Request = httptest.NewRequest(http.MethodPost, "/v1/chat/completions", nil)
|
||||
req := &dto.GeneralOpenAIRequest{
|
||||
Model: "gemini-3.1-pro-preview-thinking",
|
||||
Messages: []dto.Message{{Role: "user", Content: "hello"}},
|
||||
ExtraBody: []byte(`{"google":{"thinking_config":{"thinking_level":"low"}}}`),
|
||||
}
|
||||
info := &relaycommon.RelayInfo{
|
||||
RelayFormat: types.RelayFormatOpenAI,
|
||||
OriginModelName: "gemini-3.1-pro-preview-thinking",
|
||||
Request: req,
|
||||
ChannelMeta: &relaycommon.ChannelMeta{UpstreamModelName: "gemini-3.1-pro-preview-thinking"},
|
||||
}
|
||||
require.NoError(t, helper.ApplyReasoningModelSuffix(c, info, req))
|
||||
assert.Equal(t, "gemini-3.1-pro-preview", info.UpstreamModelName)
|
||||
|
||||
result, err := service.ConvertRequest(c, info, types.RelayFormatGemini, req)
|
||||
require.NoError(t, err)
|
||||
converted, ok := result.Value.(*dto.GeminiChatRequest)
|
||||
require.True(t, ok)
|
||||
require.NotNil(t, converted.GenerationConfig.ThinkingConfig)
|
||||
assert.Equal(t, "low", converted.GenerationConfig.ThinkingConfig.ThinkingLevel)
|
||||
assert.Nil(t, converted.GenerationConfig.ThinkingConfig.ThinkingBudget)
|
||||
})
|
||||
|
||||
t.Run("gemini thinkingBudget converts to openai reasoning_effort", func(t *testing.T) {
|
||||
settings.ThinkingAdapterEnabled = false
|
||||
c, _ := gin.CreateTestContext(httptest.NewRecorder())
|
||||
c.Request = httptest.NewRequest(http.MethodPost, "/v1beta/models/gemini-3.1-pro-preview:generateContent", nil)
|
||||
budget := 8192
|
||||
req := &dto.GeminiChatRequest{
|
||||
Contents: []dto.GeminiChatContent{{Role: "user", Parts: []dto.GeminiPart{{Text: "hello"}}}},
|
||||
GenerationConfig: dto.GeminiChatGenerationConfig{
|
||||
ThinkingConfig: &dto.GeminiThinkingConfig{ThinkingBudget: &budget},
|
||||
},
|
||||
}
|
||||
info := &relaycommon.RelayInfo{
|
||||
RelayFormat: types.RelayFormatGemini,
|
||||
OriginModelName: "gemini-3.1-pro-preview",
|
||||
Request: req,
|
||||
ChannelMeta: &relaycommon.ChannelMeta{UpstreamModelName: "gpt-5"},
|
||||
}
|
||||
|
||||
result, err := service.ConvertRequest(c, info, types.RelayFormatOpenAI, req)
|
||||
require.NoError(t, err)
|
||||
converted, ok := result.Value.(*dto.GeneralOpenAIRequest)
|
||||
require.True(t, ok)
|
||||
assert.Equal(t, "medium", converted.ReasoningEffort)
|
||||
assert.True(t, hasHostDiagnosticCode(info.ConversionDiagnostics(), "gemini_budget_to_level"))
|
||||
})
|
||||
}
|
||||
Reference in new issue
Block a user