Files
new-api/relay/channel/task/vertex/adaptor.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

418 lines
13 KiB
Go

package vertex
import (
"bytes"
"fmt"
"io"
"net/http"
"regexp"
"strings"
"time"
"github.com/QuantumNous/new-api/common"
"github.com/QuantumNous/new-api/model"
"github.com/gin-gonic/gin"
"github.com/QuantumNous/new-api/constant"
taskdto "github.com/QuantumNous/new-api/dto"
"github.com/QuantumNous/new-api/relay/channel"
geminitask "github.com/QuantumNous/new-api/relay/channel/task/gemini"
taskcommon "github.com/QuantumNous/new-api/relay/channel/task/taskcommon"
vertexcore "github.com/QuantumNous/new-api/relay/channel/vertex"
relaycommon "github.com/QuantumNous/new-api/relay/common"
"github.com/QuantumNous/new-api/relaykit/dto"
"github.com/QuantumNous/new-api/service"
)
// ============================
// Request / Response structures
// ============================
type fetchOperationPayload struct {
OperationName string `json:"operationName"`
}
type submitResponse struct {
Name string `json:"name"`
}
type operationVideo struct {
MimeType string `json:"mimeType"`
BytesBase64Encoded string `json:"bytesBase64Encoded"`
Encoding string `json:"encoding"`
}
type operationResponse struct {
Name string `json:"name"`
Done bool `json:"done"`
Response struct {
Type string `json:"@type"`
RaiMediaFilteredCount int `json:"raiMediaFilteredCount"`
Videos []operationVideo `json:"videos"`
BytesBase64Encoded string `json:"bytesBase64Encoded"`
Encoding string `json:"encoding"`
Video string `json:"video"`
} `json:"response"`
Error struct {
Message string `json:"message"`
} `json:"error"`
}
// ============================
// Adaptor implementation
// ============================
type TaskAdaptor struct {
taskcommon.BaseBilling
ChannelType int
apiKey string
baseURL string
}
func (a *TaskAdaptor) Init(info *relaycommon.RelayInfo) {
a.ChannelType = info.ChannelType
a.baseURL = info.ChannelBaseUrl
a.apiKey = info.ApiKey
}
// ValidateRequestAndSetAction parses body, validates fields and sets default action.
func (a *TaskAdaptor) ValidateRequestAndSetAction(c *gin.Context, info *relaycommon.RelayInfo) (taskErr *taskdto.TaskError) {
// Use the standard validation method for TaskSubmitReq
return relaycommon.ValidateBasicTaskRequest(c, info, constant.TaskActionTextGenerate)
}
// BuildRequestURL constructs the upstream URL.
func (a *TaskAdaptor) BuildRequestURL(info *relaycommon.RelayInfo) (string, error) {
adc := &vertexcore.Credentials{}
if err := common.Unmarshal([]byte(a.apiKey), adc); err != nil {
return "", fmt.Errorf("failed to decode credentials: %w", err)
}
modelName := info.UpstreamModelName
if modelName == "" {
modelName = "veo-3.0-generate-001"
}
region := vertexcore.GetModelRegion(info.ApiVersion, modelName)
if strings.TrimSpace(region) == "" {
region = "global"
}
return vertexcore.BuildGoogleModelURL(a.baseURL, vertexcore.DefaultAPIVersion, adc.ProjectID, region, modelName, "predictLongRunning"), nil
}
// BuildRequestHeader sets required headers.
func (a *TaskAdaptor) BuildRequestHeader(c *gin.Context, req *http.Request, info *relaycommon.RelayInfo) error {
req.Header.Set("Content-Type", "application/json")
req.Header.Set("Accept", "application/json")
adc := &vertexcore.Credentials{}
if err := common.Unmarshal([]byte(a.apiKey), adc); err != nil {
return fmt.Errorf("failed to decode credentials: %w", err)
}
proxy := ""
if info != nil {
proxy = info.ChannelSetting.Proxy
}
token, err := vertexcore.AcquireAccessToken(*adc, proxy)
if err != nil {
return fmt.Errorf("failed to acquire access token: %w", err)
}
req.Header.Set("Authorization", "Bearer "+token)
req.Header.Set("x-goog-user-project", adc.ProjectID)
return nil
}
// EstimateBilling returns OtherRatios based on durationSeconds and resolution.
func (a *TaskAdaptor) EstimateBilling(c *gin.Context, info *relaycommon.RelayInfo) map[string]float64 {
v, ok := c.Get("task_request")
if !ok {
return nil
}
req := v.(relaycommon.TaskSubmitReq)
seconds := geminitask.ResolveVeoDuration(req.Metadata, req.Duration, req.Seconds)
resolution := geminitask.ResolveVeoResolution(req.Metadata, req.Size)
resRatio := geminitask.VeoResolutionRatio(info.UpstreamModelName, resolution)
return map[string]float64{
"seconds": float64(seconds),
"resolution": resRatio,
}
}
// BuildRequestBody converts request into Vertex specific format.
func (a *TaskAdaptor) BuildRequestBody(c *gin.Context, info *relaycommon.RelayInfo) (io.Reader, error) {
v, ok := c.Get("task_request")
if !ok {
return nil, fmt.Errorf("request not found in context")
}
req := v.(relaycommon.TaskSubmitReq)
instance := geminitask.VeoInstance{Prompt: req.Prompt}
if img := geminitask.ExtractMultipartImage(c, info); img != nil {
instance.Image = img
} else if len(req.Images) > 0 {
if parsed := geminitask.ParseImageInput(req.Images[0]); parsed != nil {
instance.Image = parsed
info.Action = constant.TaskActionGenerate
}
}
params := &geminitask.VeoParameters{}
if err := taskcommon.UnmarshalMetadata(req.Metadata, params); err != nil {
return nil, fmt.Errorf("unmarshal metadata failed: %w", err)
}
if params.DurationSeconds == 0 && req.Duration > 0 {
params.DurationSeconds = req.Duration
}
if params.Resolution == "" && req.Size != "" {
params.Resolution = geminitask.SizeToVeoResolution(req.Size)
}
if params.AspectRatio == "" && req.Size != "" {
params.AspectRatio = geminitask.SizeToVeoAspectRatio(req.Size)
}
params.Resolution = strings.ToLower(params.Resolution)
params.SampleCount = 1
body := geminitask.VeoRequestPayload{
Instances: []geminitask.VeoInstance{instance},
Parameters: params,
}
data, err := common.Marshal(body)
if err != nil {
return nil, err
}
return bytes.NewReader(data), nil
}
// DoRequest delegates to common helper.
func (a *TaskAdaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
return channel.DoTaskApiRequest(a, c, info, requestBody)
}
// DoResponse handles upstream response, returns taskID etc.
func (a *TaskAdaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (taskID string, taskData []byte, taskErr *taskdto.TaskError) {
responseBody, err := io.ReadAll(resp.Body)
if err != nil {
return "", nil, service.TaskErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError)
}
_ = resp.Body.Close()
var s submitResponse
if err := common.Unmarshal(responseBody, &s); err != nil {
return "", nil, service.TaskErrorWrapper(err, "unmarshal_response_failed", http.StatusInternalServerError)
}
if strings.TrimSpace(s.Name) == "" {
return "", nil, service.TaskErrorWrapper(fmt.Errorf("missing operation name"), "invalid_response", http.StatusInternalServerError)
}
localID := taskcommon.EncodeLocalTaskID(s.Name)
ov := dto.NewOpenAIVideo()
ov.ID = info.PublicTaskID
ov.TaskID = info.PublicTaskID
ov.CreatedAt = time.Now().Unix()
ov.Model = info.OriginModelName
c.JSON(http.StatusOK, ov)
return localID, responseBody, nil
}
func (a *TaskAdaptor) GetModelList() []string {
return []string{
"veo-3.0-generate-001",
"veo-3.0-fast-generate-001",
"veo-3.1-generate-preview",
"veo-3.1-fast-generate-preview",
}
}
func (a *TaskAdaptor) GetChannelName() string { return "vertex" }
func buildFetchOperationURL(baseURL, upstreamName string) (string, error) {
region := extractRegionFromOperationName(upstreamName)
if region == "" {
region = "us-central1"
}
project := extractProjectFromOperationName(upstreamName)
modelName := extractModelFromOperationName(upstreamName)
if strings.TrimSpace(modelName) == "" {
return "", fmt.Errorf("cannot extract model from operation name")
}
if strings.TrimSpace(project) == "" {
return "", fmt.Errorf("cannot extract project from operation name")
}
return vertexcore.BuildGoogleModelURL(baseURL, vertexcore.DefaultAPIVersion, project, region, modelName, "fetchPredictOperation"), nil
}
// FetchTask fetch task status
func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
taskID, ok := body["task_id"].(string)
if !ok {
return nil, fmt.Errorf("invalid task_id")
}
upstreamName, err := taskcommon.DecodeLocalTaskID(taskID)
if err != nil {
return nil, fmt.Errorf("decode task_id failed: %w", err)
}
url, err := buildFetchOperationURL(baseUrl, upstreamName)
if err != nil {
return nil, err
}
payload := fetchOperationPayload{OperationName: upstreamName}
data, err := common.Marshal(payload)
if err != nil {
return nil, err
}
adc := &vertexcore.Credentials{}
if err := common.Unmarshal([]byte(key), adc); err != nil {
return nil, fmt.Errorf("failed to decode credentials: %w", err)
}
token, err := vertexcore.AcquireAccessToken(*adc, proxy)
if err != nil {
return nil, fmt.Errorf("failed to acquire access token: %w", err)
}
req, err := http.NewRequest(http.MethodPost, url, bytes.NewReader(data))
if err != nil {
return nil, err
}
req.Header.Set("Content-Type", "application/json")
req.Header.Set("Accept", "application/json")
req.Header.Set("Authorization", "Bearer "+token)
req.Header.Set("x-goog-user-project", adc.ProjectID)
client, err := service.GetHttpClientWithProxy(proxy)
if err != nil {
return nil, fmt.Errorf("new proxy http client failed: %w", err)
}
return client.Do(req)
}
func (a *TaskAdaptor) ParseTaskResult(respBody []byte) (*relaycommon.TaskInfo, error) {
var op operationResponse
if err := common.Unmarshal(respBody, &op); err != nil {
return nil, fmt.Errorf("unmarshal operation response failed: %w", err)
}
ti := &relaycommon.TaskInfo{}
if op.Error.Message != "" {
ti.Status = model.TaskStatusFailure
ti.Reason = op.Error.Message
ti.Progress = "100%"
return ti, nil
}
if !op.Done {
ti.Status = model.TaskStatusInProgress
ti.Progress = "50%"
return ti, nil
}
ti.Status = model.TaskStatusSuccess
ti.Progress = "100%"
if len(op.Response.Videos) > 0 {
v0 := op.Response.Videos[0]
if v0.BytesBase64Encoded != "" {
mime := strings.TrimSpace(v0.MimeType)
if mime == "" {
enc := strings.TrimSpace(v0.Encoding)
if enc == "" {
enc = "mp4"
}
if strings.Contains(enc, "/") {
mime = enc
} else {
mime = "video/" + enc
}
}
ti.Url = "data:" + mime + ";base64," + v0.BytesBase64Encoded
return ti, nil
}
}
if op.Response.BytesBase64Encoded != "" {
enc := strings.TrimSpace(op.Response.Encoding)
if enc == "" {
enc = "mp4"
}
mime := enc
if !strings.Contains(enc, "/") {
mime = "video/" + enc
}
ti.Url = "data:" + mime + ";base64," + op.Response.BytesBase64Encoded
return ti, nil
}
if op.Response.Video != "" { // some variants use `video` as base64
enc := strings.TrimSpace(op.Response.Encoding)
if enc == "" {
enc = "mp4"
}
mime := enc
if !strings.Contains(enc, "/") {
mime = "video/" + enc
}
ti.Url = "data:" + mime + ";base64," + op.Response.Video
return ti, nil
}
return ti, nil
}
func (a *TaskAdaptor) ConvertToOpenAIVideo(task *model.Task) ([]byte, error) {
// Use GetUpstreamTaskID() to get the real upstream operation name for model extraction.
// task.TaskID is now a public task_xxxx ID, no longer a base64-encoded upstream name.
upstreamTaskID := task.GetUpstreamTaskID()
upstreamName, err := taskcommon.DecodeLocalTaskID(upstreamTaskID)
if err != nil {
upstreamName = ""
}
modelName := extractModelFromOperationName(upstreamName)
if strings.TrimSpace(modelName) == "" {
modelName = "veo-3.0-generate-001"
}
v := dto.NewOpenAIVideo()
v.ID = task.TaskID
v.Model = modelName
v.Status = task.Status.ToVideoStatus()
v.SetProgressStr(task.Progress)
v.CreatedAt = task.CreatedAt
v.CompletedAt = task.UpdatedAt
if resultURL := task.GetResultURL(); strings.HasPrefix(resultURL, "data:") && len(resultURL) > 0 {
v.SetMetadata("url", resultURL)
}
return common.Marshal(v)
}
// ============================
// helpers
// ============================
var regionRe = regexp.MustCompile(`locations/([a-z0-9-]+)/`)
func extractRegionFromOperationName(name string) string {
m := regionRe.FindStringSubmatch(name)
if len(m) == 2 {
return m[1]
}
return ""
}
var modelRe = regexp.MustCompile(`models/([^/]+)/operations/`)
func extractModelFromOperationName(name string) string {
m := modelRe.FindStringSubmatch(name)
if len(m) == 2 {
return m[1]
}
idx := strings.Index(name, "models/")
if idx >= 0 {
s := name[idx+len("models/"):]
if p := strings.Index(s, "/operations/"); p > 0 {
return s[:p]
}
}
return ""
}
var projectRe = regexp.MustCompile(`projects/([^/]+)/locations/`)
func extractProjectFromOperationName(name string) string {
m := projectRe.FindStringSubmatch(name)
if len(m) == 2 {
return m[1]
}
return ""
}