Files
new-api/relaykit/relayconvert/internal/oai_responses/to_gemini_chat_req.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

310 lines
9.2 KiB
Go

package oairesponses
import (
"fmt"
"strings"
"context"
"github.com/QuantumNous/new-api/relaykit/dto"
"github.com/QuantumNous/new-api/relaykit/relayconvert/convmeta"
relaymedia "github.com/QuantumNous/new-api/relaykit/relayconvert/internal/media"
sharedgemini "github.com/QuantumNous/new-api/relaykit/relayconvert/internal/shared/gemini"
kitutil "github.com/QuantumNous/new-api/relaykit/relayconvert/kitutil"
)
func convertOpenAIResponsesRequestToGeminiChat(c context.Context, info convmeta.Meta, request any) (any, error) {
responsesRequest, err := OpenAIResponsesRequestFromAny(request)
if err != nil {
return nil, err
}
prepared, err := PrepareOpenAIResponsesRequest(*responsesRequest)
if err != nil {
return nil, err
}
return OpenAIResponsesRequestToGeminiChat(c, &prepared, info)
}
func OpenAIResponsesRequestToGeminiChat(c context.Context, req *dto.OpenAIResponsesRequest, info convmeta.Meta) (*dto.GeminiChatRequest, error) {
opts := convmeta.OptionsOf(info)
if req == nil {
return nil, fmt.Errorf("request is nil")
}
if req.Model == "" {
return nil, fmt.Errorf("model is required")
}
if err := ValidateRequestChatUnsupportedFields(req); err != nil {
return nil, err
}
geminiRequest := &dto.GeminiChatRequest{
GenerationConfig: dto.GeminiChatGenerationConfig{
Temperature: req.Temperature,
},
}
if req.TopP != nil && *req.TopP > 0 {
geminiRequest.GenerationConfig.TopP = kitutil.GetPointer(*req.TopP)
}
if req.MaxOutputTokens != nil && *req.MaxOutputTokens > 0 {
geminiRequest.GenerationConfig.MaxOutputTokens = kitutil.GetPointer(*req.MaxOutputTokens)
}
upstreamModelName := req.Model
if modelName := convmeta.UpstreamModelName(info); modelName != "" {
upstreamModelName = modelName
}
if opts.Gemini.SupportsImagineModel(upstreamModelName) {
geminiRequest.GenerationConfig.ResponseModalities = []string{"TEXT", "IMAGE"}
}
if err := applyResponsesTextToGemini(req.Text, geminiRequest); err != nil {
return nil, err
}
sharedgemini.ApplyThinkingConfig(geminiRequest, info, dto.GeneralOpenAIRequest{
Model: req.Model,
MaxCompletionTokens: req.MaxOutputTokens,
ReasoningEffort: ReasoningEffort(req),
})
var safetySettings []dto.GeminiChatSafetySettings
for _, category := range sharedgemini.SafetySettingCategories {
threshold := opts.Gemini.SafetySettingFor(category)
if threshold == "" {
continue
}
safetySettings = append(safetySettings, dto.GeminiChatSafetySettings{
Category: category,
Threshold: threshold,
})
}
if len(safetySettings) > 0 {
geminiRequest.SafetySettings = safetySettings
}
functions, err := RequestFunctionDeclarations(req.Tools)
if err != nil {
return nil, err
}
for i := range functions {
if params, ok := functions[i].Parameters.(map[string]interface{}); ok {
if props, hasProps := params["properties"].(map[string]interface{}); hasProps && len(props) == 0 {
functions[i].Parameters = nil
continue
}
}
functions[i].Parameters = sharedgemini.CleanFunctionParameters(functions[i].Parameters)
}
if len(functions) > 0 {
geminiRequest.SetTools([]dto.GeminiChatTool{
{FunctionDeclarations: functions},
})
}
toolChoice, err := RequestToolChoiceToChat(req.ToolChoice)
if err != nil {
return nil, err
}
if toolChoice != nil {
geminiRequest.ToolConfig = sharedgemini.OpenAIToolChoiceToConfig(toolChoice)
}
systemTexts := make([]string, 0)
if RawJSONPresent(req.Instructions) {
instructions, err := JSONString(req.Instructions)
if err != nil {
return nil, fmt.Errorf("invalid instructions: %w", err)
}
if strings.TrimSpace(instructions) != "" {
systemTexts = append(systemTexts, instructions)
}
}
inputItems, err := InputItems(req.Input)
if err != nil {
return nil, err
}
callNames := make(map[string]string)
for _, item := range inputItems {
itemType := strings.TrimSpace(kitutil.Interface2String(item["type"]))
switch itemType {
case ResponsesInputTypeFunctionCall:
part, callID, err := responsesFunctionCallItemToGeminiPart(item)
if err != nil {
return nil, err
}
sharedgemini.AttachFunctionCallThoughtSignature(opts, &part)
if callID != "" {
callNames[callID] = part.FunctionCall.FunctionName
}
appendGeminiContentPart(geminiRequest, "model", part)
case ResponsesInputTypeFunctionCallOutput:
part := responsesFunctionOutputItemToGeminiPart(item, callNames)
appendGeminiContentPart(geminiRequest, "user", part)
default:
role := responsesGeminiRole(item)
parts, err := responsesInputContentToGeminiParts(c, item["content"])
if err != nil {
return nil, err
}
if role == "system" {
for _, part := range parts {
if part.Text != "" {
systemTexts = append(systemTexts, part.Text)
}
}
continue
}
if len(parts) > 0 {
geminiRequest.Contents = append(geminiRequest.Contents, dto.GeminiChatContent{
Role: role,
Parts: parts,
})
}
}
}
if len(systemTexts) > 0 {
geminiRequest.SystemInstructions = &dto.GeminiChatContent{
Parts: []dto.GeminiPart{{Text: strings.Join(systemTexts, "\n")}},
}
}
return geminiRequest, nil
}
func applyResponsesTextToGemini(raw []byte, geminiRequest *dto.GeminiChatRequest) error {
responseFormat, err := RequestTextToChatResponseFormat(raw)
if err != nil {
return err
}
if responseFormat == nil || (responseFormat.Type != "json_schema" && responseFormat.Type != "json_object") {
return nil
}
geminiRequest.GenerationConfig.ResponseMimeType = "application/json"
if len(responseFormat.JsonSchema) == 0 {
return nil
}
var jsonSchema dto.FormatJsonSchema
if err := kitutil.Unmarshal(responseFormat.JsonSchema, &jsonSchema); err != nil {
return nil
}
geminiRequest.GenerationConfig.ResponseSchema = sharedgemini.RemoveAdditionalProperties(jsonSchema.Schema, 0)
return nil
}
func responsesInputContentToGeminiParts(c context.Context, content any) ([]dto.GeminiPart, error) {
contentParts, err := ContentParts(content)
if err != nil {
return nil, err
}
parts := make([]dto.GeminiPart, 0, len(contentParts))
for _, contentPart := range contentParts {
nextParts, err := responsesContentPartToGeminiParts(c, contentPart)
if err != nil {
return nil, err
}
parts = append(parts, nextParts...)
}
return parts, nil
}
func responsesContentPartToGeminiParts(c context.Context, part map[string]any) ([]dto.GeminiPart, error) {
partType := strings.TrimSpace(kitutil.Interface2String(part["type"]))
switch partType {
case "input_text", "output_text", "text":
text := kitutil.Interface2String(part["text"])
if text == "" {
return nil, nil
}
return []dto.GeminiPart{{Text: text}}, nil
case "input_image", "input_file", "input_audio", "input_video":
source := ContentPartToFileSource(part)
if source == nil {
return nil, nil
}
base64Data, mimeType, err := relaymedia.ResolveBase64Data(c, source, "formatting Responses input for Gemini")
if err != nil {
return nil, fmt.Errorf("get file data from '%s' failed: %w", source.GetIdentifier(), err)
}
if _, ok := sharedgemini.SupportedMimeTypes[strings.ToLower(mimeType)]; !ok {
return nil, fmt.Errorf("mime type is not supported by Gemini: '%s', url: '%s', supported types are: %v", mimeType, source.GetIdentifier(), sharedgemini.SupportedMimeTypesList())
}
return []dto.GeminiPart{
{
InlineData: &dto.GeminiInlineData{
MimeType: mimeType,
Data: base64Data,
},
},
}, nil
default:
return nil, nil
}
}
func responsesFunctionCallItemToGeminiPart(item map[string]any) (dto.GeminiPart, string, error) {
name := strings.TrimSpace(kitutil.Interface2String(item["name"]))
if name == "" {
return dto.GeminiPart{}, "", fmt.Errorf("function_call item is missing name")
}
callID := CallID(item)
return dto.GeminiPart{
FunctionCall: &dto.FunctionCall{
FunctionName: name,
Arguments: ObjectValue(item["arguments"], "arguments"),
},
}, callID, nil
}
func responsesFunctionOutputItemToGeminiPart(item map[string]any, callNames map[string]string) dto.GeminiPart {
callID := CallID(item)
name := strings.TrimSpace(kitutil.Interface2String(item["name"]))
if name == "" {
name = callNames[callID]
}
return dto.GeminiPart{
FunctionResponse: &dto.GeminiFunctionResponse{
Name: name,
Response: GeminiResponseMap(item["output"]),
},
}
}
func appendGeminiContentPart(req *dto.GeminiChatRequest, role string, part dto.GeminiPart) {
if len(req.Contents) > 0 && req.Contents[len(req.Contents)-1].Role == role {
if role == "model" && part.FunctionCall != nil {
parts := req.Contents[len(req.Contents)-1].Parts
insertAt := 0
for insertAt < len(parts) && parts[insertAt].FunctionCall != nil {
insertAt++
}
parts = append(parts, dto.GeminiPart{})
copy(parts[insertAt+1:], parts[insertAt:])
parts[insertAt] = part
req.Contents[len(req.Contents)-1].Parts = parts
return
}
req.Contents[len(req.Contents)-1].Parts = append(req.Contents[len(req.Contents)-1].Parts, part)
return
}
req.Contents = append(req.Contents, dto.GeminiChatContent{
Role: role,
Parts: []dto.GeminiPart{part},
})
}
func responsesGeminiRole(item map[string]any) string {
switch strings.TrimSpace(kitutil.Interface2String(item["role"])) {
case "assistant":
return "model"
case "system", "developer":
return "system"
case "model":
return "model"
default:
return "user"
}
}