Files
new-api/relaykit/relayconvert/internal/gemini_chat/to_oai_chat_resp.go
T
Calcium-Ion 86ac0f7745 refactor: extract protocol conversion layer into standalone relaykit module (#6369)
* test(relayconvert): add golden snapshot matrix and relaykit boundary guard

Phase 0 of the relaykit extraction plan: pin byte-level output of every
registered (from,to) request/response/stream conversion route, and
forbid kit-bound packages from growing host-only imports.

* wip(relayconvert): drop gin.Context from converter signatures; add convmeta draft

Phase 1 in progress: relayconvert now takes context.Context; host media
resolver adapts gin.Context back at the service boundary.

* refactor(relayconvert): decouple converters from RelayInfo, gin, and settings

Phase 1 of the relaykit extraction plan:
- converters now depend on convmeta.Meta (implemented by RelayInfo) instead
  of *relaycommon.RelayInfo; ClaudeConvertInfo and the format guesser move
  to convmeta with aliases left behind
- host settings reach converters via a convmeta.Options snapshot built in
  RelayInfo.ConvOptions; no more model_setting/reasoning global reads inside
  the conversion layer
- effort-suffix helpers move to service/relayconvert/reasoning (old package
  forwards); chat-to-responses upgrade policy moves to service (host routing
  logic, not conversion)
- golden conversion matrix unchanged

* test(relayconvert): tighten boundary — kit packages now free of gin/setting imports

* refactor(dto): drop gin and logger dependencies

Phase 2 (part 1): dto.Request.IsStream now takes *http.Request instead of
*gin.Context (Gemini's impl reads query/path off the std request); dto's
three logger calls become common.SysError. Boundary test allowlist is now
empty — kit-bound packages import no gin/setting/logger/model.

* refactor(kit): extract dependency-free kitutil; dto/types/relayconvert stop importing common

Phase 2 of the relaykit extraction plan:
- new service/relayconvert/kitutil holds the pure helpers the kit needs
  (JSON wrappers, pointer/string/uuid/timestamp utils, MaskSensitiveInfo,
  pluggable LogInfo/LogError hooks, Debug flag)
- dto, types, and all relayconvert packages now use kitutil; their only
  remaining internal deps are dto/types/constant
- common keeps every original symbol (MaskSensitiveInfo delegates to
  kitutil) so host code is untouched; main.go routes kit logging into
  common.SysLog/SysError and mirrors DebugEnabled
- golden conversion matrix unchanged

* refactor(kit): move EndpointType/FinishReason to types; OpenRouter dialect via Options

Kit packages (dto/types/relayconvert/reasonmap) no longer import constant:
- EndpointType and finish-reason values live in types; constant re-exports
- the OpenRouter special-case in claude->openai request conversion reads
  Options.OpenRouterDialect, set by the host from the channel type;
  InitChannelMeta invalidates the cached snapshot on channel switch

* refactor: extract relaykit submodule (dto/types/relayconvert/reasonmap)

Phase 3 of the relaykit extraction plan:
- new go module github.com/QuantumNous/new-api/relaykit containing dto
  (minus task family), types, relayconvert (with convmeta/kitutil/reasoning),
  and reasonmap; host consumes it via require + replace, go.work for dev
- task-family dto (task/suno/midjourney/video) stays in the host dto
  package; dual-consumer host files alias it as taskdto
- relaykit builds and tests standalone (GOWORK=off): no host imports,
  no gin, no DB, no settings
- golden conversion matrix unchanged

* build(docker): copy relaykit/go.mod before go mod download

The local-replace submodule's go.mod must exist inside the build context
for the main module graph to resolve.

* fix: address relaykit extraction regressions

* fix: address relaykit review regressions

* docs: document Meta nil receiver contract

* fix(relaykit): fail OpenAI→Claude conversion without max_tokens; reject negative default_max_tokens

The Claude Messages API requires max_tokens (omitting it is a 400
"Field required"), but with a nil Options.Claude.DefaultMaxTokens hook
the converters silently emitted a request the upstream is guaranteed to
reject. Both OpenAI Chat and Responses → Claude conversions now return
sharedclaude.ErrMissingMaxTokens when no path (client value, default
hook, thinking-adapter floor) supplied one. Unreachable in the host,
which always configures the hook.

Host side, claude.default_max_tokens now rejects negative values at the
option API before persisting — they would wrap into huge unsigned values
during conversion. Zero stays allowed: the current API treats
max_tokens: 0 as cache pre-warming.

* fix: make Gemini safety settings read path race-free
2026-07-27 15:56:21 +08:00

299 lines
9.2 KiB
Go

package geminichat
import (
"fmt"
"strings"
"github.com/QuantumNous/new-api/relaykit/dto"
kitutil "github.com/QuantumNous/new-api/relaykit/relayconvert/kitutil"
"github.com/QuantumNous/new-api/relaykit/types"
)
func UsageFromGeminiMetadata(metadata *dto.GeminiUsageMetadata, fallbackPromptTokens int) *dto.Usage {
if metadata == nil {
if fallbackPromptTokens <= 0 {
return nil
}
usage := &dto.Usage{PromptTokens: fallbackPromptTokens}
usage.PromptTokensDetails.TextTokens = fallbackPromptTokens
return usage
}
promptTokens := metadata.PromptTokenCount + metadata.ToolUsePromptTokenCount
if promptTokens <= 0 && fallbackPromptTokens > 0 {
promptTokens = fallbackPromptTokens
}
usage := &dto.Usage{
PromptTokens: promptTokens,
CompletionTokens: metadata.CandidatesTokenCount + metadata.ThoughtsTokenCount,
TotalTokens: metadata.TotalTokenCount,
BillingUsage: dto.CloneBillingUsage(metadata.BillingUsage),
}
if usage.BillingUsage == nil {
usage.BillingUsage = dto.NewGeminiChatBillingUsage(metadata)
}
usage.CompletionTokenDetails.ReasoningTokens = metadata.ThoughtsTokenCount
usage.PromptTokensDetails.CachedTokens = metadata.CachedContentTokenCount
for _, detail := range metadata.PromptTokensDetails {
if detail.Modality == "AUDIO" {
usage.PromptTokensDetails.AudioTokens += detail.TokenCount
} else if detail.Modality == "IMAGE" {
usage.PromptTokensDetails.ImageTokens += detail.TokenCount
} else if detail.Modality == "TEXT" {
usage.PromptTokensDetails.TextTokens += detail.TokenCount
}
}
for _, detail := range metadata.ToolUsePromptTokensDetails {
if detail.Modality == "AUDIO" {
usage.PromptTokensDetails.AudioTokens += detail.TokenCount
} else if detail.Modality == "IMAGE" {
usage.PromptTokensDetails.ImageTokens += detail.TokenCount
} else if detail.Modality == "TEXT" {
usage.PromptTokensDetails.TextTokens += detail.TokenCount
}
}
for _, detail := range metadata.CandidatesTokensDetails {
switch detail.Modality {
case "IMAGE":
usage.CompletionTokenDetails.ImageTokens += detail.TokenCount
case "AUDIO":
usage.CompletionTokenDetails.AudioTokens += detail.TokenCount
case "TEXT":
usage.CompletionTokenDetails.TextTokens += detail.TokenCount
}
}
if usage.TotalTokens > 0 && usage.CompletionTokens <= 0 {
usage.CompletionTokens = usage.TotalTokens - usage.PromptTokens
}
if usage.PromptTokens > 0 && usage.PromptTokensDetails.TextTokens == 0 && usage.PromptTokensDetails.AudioTokens == 0 {
usage.PromptTokensDetails.TextTokens = usage.PromptTokens
}
return usage
}
func ResponseGeminiChat2OpenAI(id string, created int64, response *dto.GeminiChatResponse) *dto.OpenAITextResponse {
fullTextResponse := dto.OpenAITextResponse{
Id: id,
Object: "chat.completion",
Created: created,
Choices: make([]dto.OpenAITextResponseChoice, 0, len(response.Candidates)),
}
isToolCall := false
for _, candidate := range response.Candidates {
choice := dto.OpenAITextResponseChoice{
Index: int(candidate.Index),
Message: dto.Message{
Role: "assistant",
Content: "",
},
FinishReason: types.FinishReasonStop,
}
if len(candidate.Content.Parts) > 0 {
var content strings.Builder
var inlineGrow int
for _, part := range candidate.Content.Parts {
if part.InlineData != nil {
inlineGrow += len(part.InlineData.MimeType) + len(part.InlineData.Data) + 32
}
}
if inlineGrow > 0 {
content.Grow(inlineGrow)
}
appended := 0
writeSep := func() {
if appended > 0 {
content.WriteByte('\n')
}
appended++
}
var toolCalls []dto.ToolCallResponse
for _, part := range candidate.Content.Parts {
if part.InlineData != nil {
if strings.HasPrefix(part.InlineData.MimeType, "image") {
writeSep()
content.WriteString("![image](data:")
content.WriteString(part.InlineData.MimeType)
content.WriteString(";base64,")
content.WriteString(part.InlineData.Data)
content.WriteByte(')')
} else {
writeSep()
content.WriteString("[media](data:")
content.WriteString(part.InlineData.MimeType)
content.WriteString(";base64,")
content.WriteString(part.InlineData.Data)
content.WriteByte(')')
}
} else if part.FunctionCall != nil {
choice.FinishReason = types.FinishReasonToolCalls
if call := geminiResponseToolCall(&part); call != nil {
toolCalls = append(toolCalls, *call)
}
} else if part.Thought {
choice.Message.ReasoningContent = &part.Text
} else {
if part.ExecutableCode != nil {
writeSep()
content.WriteString("```")
content.WriteString(part.ExecutableCode.Language)
content.WriteByte('\n')
content.WriteString(part.ExecutableCode.Code)
content.WriteString("\n```")
} else if part.CodeExecutionResult != nil {
writeSep()
content.WriteString("```output\n")
content.WriteString(part.CodeExecutionResult.Output)
content.WriteString("\n```")
} else if part.Text != "\n" {
writeSep()
content.WriteString(part.Text)
}
}
}
if len(toolCalls) > 0 {
choice.Message.SetToolCalls(toolCalls)
isToolCall = true
}
choice.Message.SetStringContent(content.String())
}
if candidate.FinishReason != nil {
switch *candidate.FinishReason {
case "STOP":
choice.FinishReason = types.FinishReasonStop
case "MAX_TOKENS":
choice.FinishReason = types.FinishReasonLength
case "SAFETY", "RECITATION", "BLOCKLIST", "PROHIBITED_CONTENT", "SPII", "OTHER":
choice.FinishReason = types.FinishReasonContentFilter
default:
choice.FinishReason = types.FinishReasonContentFilter
}
}
if isToolCall {
choice.FinishReason = types.FinishReasonToolCalls
}
fullTextResponse.Choices = append(fullTextResponse.Choices, choice)
}
return &fullTextResponse
}
func StreamResponseGeminiChat2OpenAI(geminiResponse *dto.GeminiChatResponse) (*dto.ChatCompletionsStreamResponse, bool) {
choices := make([]dto.ChatCompletionsStreamResponseChoice, 0, len(geminiResponse.Candidates))
isStop := false
for _, candidate := range geminiResponse.Candidates {
if candidate.FinishReason != nil && *candidate.FinishReason == "STOP" {
isStop = true
candidate.FinishReason = nil
}
choice := dto.ChatCompletionsStreamResponseChoice{
Index: int(candidate.Index),
Delta: dto.ChatCompletionsStreamResponseChoiceDelta{},
}
var content strings.Builder
var inlineGrow int
for _, part := range candidate.Content.Parts {
if part.InlineData != nil {
inlineGrow += len(part.InlineData.MimeType) + len(part.InlineData.Data) + 32
}
}
if inlineGrow > 0 {
content.Grow(inlineGrow)
}
appended := 0
writeSep := func() {
if appended > 0 {
content.WriteByte('\n')
}
appended++
}
isTools := false
isThought := false
if candidate.FinishReason != nil {
switch *candidate.FinishReason {
case "STOP":
choice.FinishReason = &types.FinishReasonStop
case "MAX_TOKENS":
choice.FinishReason = &types.FinishReasonLength
case "SAFETY", "RECITATION", "BLOCKLIST", "PROHIBITED_CONTENT", "SPII", "OTHER":
choice.FinishReason = &types.FinishReasonContentFilter
default:
choice.FinishReason = &types.FinishReasonContentFilter
}
}
for _, part := range candidate.Content.Parts {
if part.InlineData != nil {
if strings.HasPrefix(part.InlineData.MimeType, "image") {
writeSep()
content.WriteString("![image](data:")
content.WriteString(part.InlineData.MimeType)
content.WriteString(";base64,")
content.WriteString(part.InlineData.Data)
content.WriteByte(')')
}
} else if part.FunctionCall != nil {
isTools = true
if call := geminiResponseToolCall(&part); call != nil {
call.SetIndex(len(choice.Delta.ToolCalls))
choice.Delta.ToolCalls = append(choice.Delta.ToolCalls, *call)
}
} else if part.Thought {
isThought = true
writeSep()
content.WriteString(part.Text)
} else {
if part.ExecutableCode != nil {
writeSep()
content.WriteString("```")
content.WriteString(part.ExecutableCode.Language)
content.WriteByte('\n')
content.WriteString(part.ExecutableCode.Code)
content.WriteString("\n```\n")
} else if part.CodeExecutionResult != nil {
writeSep()
content.WriteString("```output\n")
content.WriteString(part.CodeExecutionResult.Output)
content.WriteString("\n```\n")
} else if part.Text != "\n" {
writeSep()
content.WriteString(part.Text)
}
}
}
if isThought {
choice.Delta.SetReasoningContent(content.String())
} else {
choice.Delta.SetContentString(content.String())
}
if isTools {
choice.FinishReason = &types.FinishReasonToolCalls
}
choices = append(choices, choice)
}
response := dto.ChatCompletionsStreamResponse{
Object: "chat.completion.chunk",
Choices: choices,
}
return &response, isStop
}
func geminiResponseToolCall(item *dto.GeminiPart) *dto.ToolCallResponse {
argsBytes, err := kitutil.Marshal(item.FunctionCall.Arguments)
if err != nil {
return nil
}
return &dto.ToolCallResponse{
ID: fmt.Sprintf("call_%s", kitutil.GetUUID()),
Type: "function",
Function: dto.FunctionResponse{
Arguments: string(argsBytes),
Name: item.FunctionCall.FunctionName,
},
}
}