mirror of
https://github.com/QuantumNous/new-api.git
synced 2026-09-12 15:21:09 +00:00
Deferred follow-ups from the relaykit-tools review cycle, verified by live end-to-end billing tests: - billing: normalize Gemini modality keys consistently between stream merge and settlement (case/whitespace variants no longer drop independent audio/image pricing) and sum duplicate modality entries on both paths - billing: sync legacy flat Claude cache-creation fields from the CacheCreation sub-object (including zeroing) and fall back to flat fields only when the snapshot never carried a sub-object, closing a stale 1h-cache overcharge path in cascaded deployments - relay: move Chat-to-Claude and Chat-to-Gemini stream conversion state from gin.Context onto RelayInfo and reset it with SendResponseCount in InitChannelMeta, so channel retries start clean while per-request state (stream error collection, conversion diagnostics, channel chain, billing accumulators) survives - relay: Claude channel now serves Gemini-format clients (request via registry conversion, response and stream composed through the Chat pivot), removing the last unimplemented conversion direction - relaykit: recognize legacy pseudo tool names (googleSearch, codeExecution, urlContext) in the toolconv decode stage and drop the string-matching bypass in the Chat-to-Gemini converter; native Gemini tool output is restored and non-Gemini targets follow standard loss diagnostics - relaykit: attach upstream Gemini usage (with billing_usage sidecar) to intermediate stream chunks so converted Claude streams report upstream truth from message_start, and preserve the sidecar through Claude stream usage merges; billing settlement unchanged - billing: clamp negative Total-Prompt completion derivation, OR the Estimated flag across cross-dialect snapshot replacement, and fill canonical OpenAI prompt details via field-wise merge
188 lines
5.5 KiB
Go
188 lines
5.5 KiB
Go
package claude
|
|
|
|
import (
|
|
"net/http/httptest"
|
|
"testing"
|
|
|
|
"github.com/QuantumNous/new-api/common"
|
|
relaycommon "github.com/QuantumNous/new-api/relay/common"
|
|
"github.com/QuantumNous/new-api/relay/helper"
|
|
"github.com/QuantumNous/new-api/relaykit/dto"
|
|
"github.com/QuantumNous/new-api/setting/model_setting"
|
|
"github.com/gin-gonic/gin"
|
|
"github.com/stretchr/testify/assert"
|
|
"github.com/stretchr/testify/require"
|
|
)
|
|
|
|
func TestConvertClaudeRequestTreatsZeroMaxTokensAsUnset(t *testing.T) {
|
|
zero := uint(0)
|
|
req := &dto.ClaudeRequest{
|
|
Model: "claude-sonnet-4-5",
|
|
MaxTokens: &zero,
|
|
Messages: []dto.ClaudeMessage{
|
|
{Role: "user", Content: "hello"},
|
|
},
|
|
}
|
|
info := &relaycommon.RelayInfo{
|
|
ChannelMeta: &relaycommon.ChannelMeta{
|
|
UpstreamModelName: "claude-sonnet-4-5",
|
|
},
|
|
}
|
|
|
|
out, err := (&Adaptor{}).ConvertClaudeRequest(nil, info, req)
|
|
require.NoError(t, err)
|
|
converted, ok := out.(*dto.ClaudeRequest)
|
|
require.True(t, ok)
|
|
require.NotNil(t, converted.MaxTokens)
|
|
assert.Equal(t, uint(model_setting.GetClaudeSettings().GetDefaultMaxTokens(req.Model)), *converted.MaxTokens)
|
|
}
|
|
|
|
func TestConvertClaudeRequestZeroMaxTokensStillRaisesThinkingBudget(t *testing.T) {
|
|
gin.SetMode(gin.TestMode)
|
|
c, _ := gin.CreateTestContext(httptest.NewRecorder())
|
|
|
|
zero := uint(0)
|
|
original := &dto.ClaudeRequest{
|
|
Model: "claude-3-7-sonnet-thinking",
|
|
MaxTokens: &zero,
|
|
Messages: []dto.ClaudeMessage{
|
|
{Role: "user", Content: "hello"},
|
|
},
|
|
}
|
|
info := &relaycommon.RelayInfo{
|
|
OriginModelName: "claude-3-7-sonnet-thinking",
|
|
Request: original,
|
|
ChannelMeta: &relaycommon.ChannelMeta{
|
|
UpstreamModelName: "claude-3-7-sonnet-thinking",
|
|
},
|
|
}
|
|
outbound, err := common.DeepCopy(original)
|
|
require.NoError(t, err)
|
|
require.NoError(t, helper.ModelMappedHelper(c, info, outbound))
|
|
require.NoError(t, helper.ApplyReasoningModelSuffix(info, outbound))
|
|
|
|
out, err := (&Adaptor{}).ConvertClaudeRequest(nil, info, outbound)
|
|
require.NoError(t, err)
|
|
converted, ok := out.(*dto.ClaudeRequest)
|
|
require.True(t, ok)
|
|
assert.Equal(t, "claude-3-7-sonnet", converted.Model)
|
|
require.NotNil(t, converted.Thinking)
|
|
require.NotNil(t, converted.MaxTokens)
|
|
assert.Greater(t, *converted.MaxTokens, uint(1024))
|
|
}
|
|
|
|
func TestConvertClaudeRequestDoesNotOverwriteTrimmedUpstreamModelName(t *testing.T) {
|
|
req := &dto.ClaudeRequest{
|
|
Model: "claude-3-7-sonnet-thinking",
|
|
Messages: []dto.ClaudeMessage{
|
|
{Role: "user", Content: "hello"},
|
|
},
|
|
}
|
|
info := &relaycommon.RelayInfo{
|
|
ChannelMeta: &relaycommon.ChannelMeta{
|
|
UpstreamModelName: "claude-3-7-sonnet",
|
|
},
|
|
}
|
|
|
|
_, err := (&Adaptor{}).ConvertClaudeRequest(nil, info, req)
|
|
require.NoError(t, err)
|
|
assert.Equal(t, "claude-3-7-sonnet", info.UpstreamModelName)
|
|
}
|
|
|
|
func geminiToClaudeInfo() *relaycommon.RelayInfo {
|
|
return &relaycommon.RelayInfo{
|
|
OriginModelName: "claude-3-7-sonnet",
|
|
ChannelMeta: &relaycommon.ChannelMeta{
|
|
UpstreamModelName: "claude-3-7-sonnet",
|
|
},
|
|
}
|
|
}
|
|
|
|
func TestConvertGeminiRequestMapsSystemInstructionToolsAndMultimodal(t *testing.T) {
|
|
req := &dto.GeminiChatRequest{
|
|
Contents: []dto.GeminiChatContent{
|
|
{
|
|
Role: "user",
|
|
Parts: []dto.GeminiPart{
|
|
{Text: "What is in this image?"},
|
|
{InlineData: &dto.GeminiInlineData{MimeType: "image/png", Data: "aGVsbG8="}},
|
|
},
|
|
},
|
|
},
|
|
SystemInstructions: &dto.GeminiChatContent{
|
|
Parts: []dto.GeminiPart{{Text: "You are a helpful assistant."}},
|
|
},
|
|
}
|
|
req.SetTools([]dto.GeminiChatTool{
|
|
{
|
|
FunctionDeclarations: []dto.FunctionRequest{
|
|
{
|
|
Name: "lookup",
|
|
Description: "Lookup data",
|
|
Parameters: map[string]any{
|
|
"type": "object",
|
|
"properties": map[string]any{"q": map[string]any{"type": "string"}},
|
|
},
|
|
},
|
|
},
|
|
},
|
|
})
|
|
|
|
out, err := (&Adaptor{}).ConvertGeminiRequest(nil, geminiToClaudeInfo(), req)
|
|
require.NoError(t, err)
|
|
converted, ok := out.(*dto.ClaudeRequest)
|
|
require.True(t, ok)
|
|
|
|
system := converted.ParseSystem()
|
|
require.NotEmpty(t, system)
|
|
assert.Contains(t, system[0].GetText(), "You are a helpful assistant.")
|
|
require.NotEmpty(t, converted.Messages)
|
|
assert.Equal(t, "user", converted.Messages[0].Role)
|
|
|
|
blocks, parseErr := converted.Messages[0].ParseContent()
|
|
require.NoError(t, parseErr)
|
|
var foundImage bool
|
|
for _, block := range blocks {
|
|
if block.Type == "image" || (block.Source != nil && block.Source.Type == "base64") {
|
|
foundImage = true
|
|
break
|
|
}
|
|
}
|
|
assert.True(t, foundImage)
|
|
|
|
require.NotNil(t, converted.Tools)
|
|
tools, err := common.Marshal(converted.Tools)
|
|
require.NoError(t, err)
|
|
assert.Contains(t, string(tools), `"lookup"`)
|
|
require.NotNil(t, converted.MaxTokens)
|
|
assert.Greater(t, *converted.MaxTokens, uint(0))
|
|
}
|
|
|
|
func TestConvertGeminiRequestThinkingConfigUsesReasoningIntent(t *testing.T) {
|
|
budget := 1024
|
|
maxTokens := uint(4096)
|
|
req := &dto.GeminiChatRequest{
|
|
Contents: []dto.GeminiChatContent{
|
|
{Role: "user", Parts: []dto.GeminiPart{{Text: "think"}}},
|
|
},
|
|
GenerationConfig: dto.GeminiChatGenerationConfig{
|
|
MaxOutputTokens: &maxTokens,
|
|
ThinkingConfig: &dto.GeminiThinkingConfig{ThinkingBudget: &budget},
|
|
},
|
|
}
|
|
|
|
out, err := (&Adaptor{}).ConvertGeminiRequest(nil, geminiToClaudeInfo(), req)
|
|
require.NoError(t, err)
|
|
converted, ok := out.(*dto.ClaudeRequest)
|
|
require.True(t, ok)
|
|
require.NotNil(t, converted.Thinking)
|
|
assert.Equal(t, "enabled", converted.Thinking.Type)
|
|
require.NotNil(t, converted.Thinking.BudgetTokens)
|
|
assert.Equal(t, 1024, *converted.Thinking.BudgetTokens)
|
|
}
|
|
|
|
func TestConvertGeminiRequestNilRequest(t *testing.T) {
|
|
_, err := (&Adaptor{}).ConvertGeminiRequest(nil, geminiToClaudeInfo(), nil)
|
|
require.Error(t, err)
|
|
}
|