adaptor.go 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387
  1. package kling
  2. import (
  3. "bytes"
  4. "fmt"
  5. "io"
  6. "net/http"
  7. "strings"
  8. "time"
  9. "github.com/QuantumNous/new-api/common"
  10. "github.com/QuantumNous/new-api/model"
  11. "github.com/samber/lo"
  12. "github.com/gin-gonic/gin"
  13. "github.com/golang-jwt/jwt/v5"
  14. "github.com/pkg/errors"
  15. "github.com/QuantumNous/new-api/constant"
  16. "github.com/QuantumNous/new-api/dto"
  17. "github.com/QuantumNous/new-api/relay/channel"
  18. taskcommon "github.com/QuantumNous/new-api/relay/channel/task/taskcommon"
  19. relaycommon "github.com/QuantumNous/new-api/relay/common"
  20. "github.com/QuantumNous/new-api/service"
  21. )
  22. // ============================
  23. // Request / Response structures
  24. // ============================
  25. type TrajectoryPoint struct {
  26. X int `json:"x"`
  27. Y int `json:"y"`
  28. }
  29. type DynamicMask struct {
  30. Mask string `json:"mask,omitempty"`
  31. Trajectories []TrajectoryPoint `json:"trajectories,omitempty"`
  32. }
  33. type CameraConfig struct {
  34. Horizontal float64 `json:"horizontal,omitempty"`
  35. Vertical float64 `json:"vertical,omitempty"`
  36. Pan float64 `json:"pan,omitempty"`
  37. Tilt float64 `json:"tilt,omitempty"`
  38. Roll float64 `json:"roll,omitempty"`
  39. Zoom float64 `json:"zoom,omitempty"`
  40. }
  41. type CameraControl struct {
  42. Type string `json:"type,omitempty"`
  43. Config *CameraConfig `json:"config,omitempty"`
  44. }
  45. type requestPayload struct {
  46. Prompt string `json:"prompt,omitempty"`
  47. Image string `json:"image,omitempty"`
  48. ImageTail string `json:"image_tail,omitempty"`
  49. NegativePrompt string `json:"negative_prompt,omitempty"`
  50. Mode string `json:"mode,omitempty"`
  51. Duration string `json:"duration,omitempty"`
  52. AspectRatio string `json:"aspect_ratio,omitempty"`
  53. ModelName string `json:"model_name,omitempty"`
  54. Model string `json:"model,omitempty"` // Compatible with upstreams that only recognize "model"
  55. CfgScale float64 `json:"cfg_scale,omitempty"`
  56. StaticMask string `json:"static_mask,omitempty"`
  57. DynamicMasks []DynamicMask `json:"dynamic_masks,omitempty"`
  58. CameraControl *CameraControl `json:"camera_control,omitempty"`
  59. CallbackUrl string `json:"callback_url,omitempty"`
  60. ExternalTaskId string `json:"external_task_id,omitempty"`
  61. }
  62. type responsePayload struct {
  63. Code int `json:"code"`
  64. Message string `json:"message"`
  65. TaskId string `json:"task_id"`
  66. RequestId string `json:"request_id"`
  67. Data struct {
  68. TaskId string `json:"task_id"`
  69. TaskStatus string `json:"task_status"`
  70. TaskStatusMsg string `json:"task_status_msg"`
  71. TaskResult struct {
  72. Videos []struct {
  73. Id string `json:"id"`
  74. Url string `json:"url"`
  75. Duration string `json:"duration"`
  76. } `json:"videos"`
  77. } `json:"task_result"`
  78. CreatedAt int64 `json:"created_at"`
  79. UpdatedAt int64 `json:"updated_at"`
  80. } `json:"data"`
  81. }
  82. // ============================
  83. // Adaptor implementation
  84. // ============================
  85. type TaskAdaptor struct {
  86. taskcommon.BaseBilling
  87. ChannelType int
  88. apiKey string
  89. baseURL string
  90. }
  91. func (a *TaskAdaptor) Init(info *relaycommon.RelayInfo) {
  92. a.ChannelType = info.ChannelType
  93. a.baseURL = info.ChannelBaseUrl
  94. a.apiKey = info.ApiKey
  95. // apiKey format: "access_key|secret_key"
  96. }
  97. // ValidateRequestAndSetAction parses body, validates fields and sets default action.
  98. func (a *TaskAdaptor) ValidateRequestAndSetAction(c *gin.Context, info *relaycommon.RelayInfo) (taskErr *dto.TaskError) {
  99. // Use the standard validation method for TaskSubmitReq
  100. return relaycommon.ValidateBasicTaskRequest(c, info, constant.TaskActionGenerate)
  101. }
  102. // BuildRequestURL constructs the upstream URL.
  103. func (a *TaskAdaptor) BuildRequestURL(info *relaycommon.RelayInfo) (string, error) {
  104. path := lo.Ternary(info.Action == constant.TaskActionGenerate, "/v1/videos/image2video", "/v1/videos/text2video")
  105. if isNewAPIRelay(info.ApiKey) {
  106. return fmt.Sprintf("%s/kling%s", a.baseURL, path), nil
  107. }
  108. return fmt.Sprintf("%s%s", a.baseURL, path), nil
  109. }
  110. // BuildRequestHeader sets required headers.
  111. func (a *TaskAdaptor) BuildRequestHeader(c *gin.Context, req *http.Request, info *relaycommon.RelayInfo) error {
  112. token, err := a.createJWTToken()
  113. if err != nil {
  114. return fmt.Errorf("failed to create JWT token: %w", err)
  115. }
  116. req.Header.Set("Content-Type", "application/json")
  117. req.Header.Set("Accept", "application/json")
  118. req.Header.Set("Authorization", "Bearer "+token)
  119. req.Header.Set("User-Agent", "kling-sdk/1.0")
  120. return nil
  121. }
  122. // BuildRequestBody converts request into Kling specific format.
  123. func (a *TaskAdaptor) BuildRequestBody(c *gin.Context, info *relaycommon.RelayInfo) (io.Reader, error) {
  124. v, exists := c.Get("task_request")
  125. if !exists {
  126. return nil, fmt.Errorf("request not found in context")
  127. }
  128. req := v.(relaycommon.TaskSubmitReq)
  129. body, err := a.convertToRequestPayload(&req, info)
  130. if err != nil {
  131. return nil, err
  132. }
  133. if body.Image == "" && body.ImageTail == "" {
  134. c.Set("action", constant.TaskActionTextGenerate)
  135. }
  136. data, err := common.Marshal(body)
  137. if err != nil {
  138. return nil, err
  139. }
  140. return bytes.NewReader(data), nil
  141. }
  142. // DoRequest delegates to common helper.
  143. func (a *TaskAdaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
  144. if action := c.GetString("action"); action != "" {
  145. info.Action = action
  146. }
  147. return channel.DoTaskApiRequest(a, c, info, requestBody)
  148. }
  149. // DoResponse handles upstream response, returns taskID etc.
  150. func (a *TaskAdaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (taskID string, taskData []byte, taskErr *dto.TaskError) {
  151. responseBody, err := io.ReadAll(resp.Body)
  152. if err != nil {
  153. taskErr = service.TaskErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError)
  154. return
  155. }
  156. var kResp responsePayload
  157. err = common.Unmarshal(responseBody, &kResp)
  158. if err != nil {
  159. taskErr = service.TaskErrorWrapper(err, "unmarshal_response_failed", http.StatusInternalServerError)
  160. return
  161. }
  162. if kResp.Code != 0 {
  163. taskErr = service.TaskErrorWrapperLocal(fmt.Errorf("%s", kResp.Message), "task_failed", http.StatusBadRequest)
  164. return
  165. }
  166. ov := dto.NewOpenAIVideo()
  167. ov.ID = info.PublicTaskID
  168. ov.TaskID = info.PublicTaskID
  169. ov.CreatedAt = time.Now().Unix()
  170. ov.Model = info.OriginModelName
  171. c.JSON(http.StatusOK, ov)
  172. return kResp.Data.TaskId, responseBody, nil
  173. }
  174. // FetchTask fetch task status
  175. func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
  176. taskID, ok := body["task_id"].(string)
  177. if !ok {
  178. return nil, fmt.Errorf("invalid task_id")
  179. }
  180. action, ok := body["action"].(string)
  181. if !ok {
  182. return nil, fmt.Errorf("invalid action")
  183. }
  184. path := lo.Ternary(action == constant.TaskActionGenerate, "/v1/videos/image2video", "/v1/videos/text2video")
  185. url := fmt.Sprintf("%s%s/%s", baseUrl, path, taskID)
  186. if isNewAPIRelay(key) {
  187. url = fmt.Sprintf("%s/kling%s/%s", baseUrl, path, taskID)
  188. }
  189. req, err := http.NewRequest(http.MethodGet, url, nil)
  190. if err != nil {
  191. return nil, err
  192. }
  193. token, err := a.createJWTTokenWithKey(key)
  194. if err != nil {
  195. token = key
  196. }
  197. req.Header.Set("Accept", "application/json")
  198. req.Header.Set("Authorization", "Bearer "+token)
  199. req.Header.Set("User-Agent", "kling-sdk/1.0")
  200. client, err := service.GetHttpClientWithProxy(proxy)
  201. if err != nil {
  202. return nil, fmt.Errorf("new proxy http client failed: %w", err)
  203. }
  204. return client.Do(req)
  205. }
  206. func (a *TaskAdaptor) GetModelList() []string {
  207. return []string{"kling-v1", "kling-v1-6", "kling-v2-master"}
  208. }
  209. func (a *TaskAdaptor) GetChannelName() string {
  210. return "kling"
  211. }
  212. // ============================
  213. // helpers
  214. // ============================
  215. func (a *TaskAdaptor) convertToRequestPayload(req *relaycommon.TaskSubmitReq, info *relaycommon.RelayInfo) (*requestPayload, error) {
  216. r := requestPayload{
  217. Prompt: req.Prompt,
  218. Image: req.Image,
  219. Mode: taskcommon.DefaultString(req.Mode, "std"),
  220. Duration: fmt.Sprintf("%d", taskcommon.DefaultInt(req.Duration, 5)),
  221. AspectRatio: a.getAspectRatio(req.Size),
  222. ModelName: info.UpstreamModelName,
  223. Model: info.UpstreamModelName,
  224. CfgScale: 0.5,
  225. StaticMask: "",
  226. DynamicMasks: []DynamicMask{},
  227. CameraControl: nil,
  228. CallbackUrl: "",
  229. ExternalTaskId: "",
  230. }
  231. if r.ModelName == "" {
  232. r.ModelName = "kling-v1"
  233. r.Model = "kling-v1"
  234. }
  235. if err := taskcommon.UnmarshalMetadata(req.Metadata, &r); err != nil {
  236. return nil, errors.Wrap(err, "unmarshal metadata failed")
  237. }
  238. return &r, nil
  239. }
  240. func (a *TaskAdaptor) getAspectRatio(size string) string {
  241. switch size {
  242. case "1024x1024", "512x512":
  243. return "1:1"
  244. case "1280x720", "1920x1080":
  245. return "16:9"
  246. case "720x1280", "1080x1920":
  247. return "9:16"
  248. default:
  249. return "1:1"
  250. }
  251. }
  252. // ============================
  253. // JWT helpers
  254. // ============================
  255. func (a *TaskAdaptor) createJWTToken() (string, error) {
  256. return a.createJWTTokenWithKey(a.apiKey)
  257. }
  258. func (a *TaskAdaptor) createJWTTokenWithKey(apiKey string) (string, error) {
  259. if isNewAPIRelay(apiKey) {
  260. return apiKey, nil // new api relay
  261. }
  262. keyParts := strings.Split(apiKey, "|")
  263. if len(keyParts) != 2 {
  264. return "", errors.New("invalid api_key, required format is accessKey|secretKey")
  265. }
  266. accessKey := strings.TrimSpace(keyParts[0])
  267. if len(keyParts) == 1 {
  268. return accessKey, nil
  269. }
  270. secretKey := strings.TrimSpace(keyParts[1])
  271. now := time.Now().Unix()
  272. claims := jwt.MapClaims{
  273. "iss": accessKey,
  274. "exp": now + 1800, // 30 minutes
  275. "nbf": now - 5,
  276. }
  277. token := jwt.NewWithClaims(jwt.SigningMethodHS256, claims)
  278. token.Header["typ"] = "JWT"
  279. return token.SignedString([]byte(secretKey))
  280. }
  281. func (a *TaskAdaptor) ParseTaskResult(respBody []byte) (*relaycommon.TaskInfo, error) {
  282. taskInfo := &relaycommon.TaskInfo{}
  283. resPayload := responsePayload{}
  284. err := common.Unmarshal(respBody, &resPayload)
  285. if err != nil {
  286. return nil, errors.Wrap(err, "failed to unmarshal response body")
  287. }
  288. taskInfo.Code = resPayload.Code
  289. taskInfo.TaskID = resPayload.Data.TaskId
  290. taskInfo.Reason = resPayload.Data.TaskStatusMsg
  291. //任务状态,枚举值:submitted(已提交)、processing(处理中)、succeed(成功)、failed(失败)
  292. status := resPayload.Data.TaskStatus
  293. switch status {
  294. case "submitted":
  295. taskInfo.Status = model.TaskStatusSubmitted
  296. case "processing":
  297. taskInfo.Status = model.TaskStatusInProgress
  298. case "succeed":
  299. taskInfo.Status = model.TaskStatusSuccess
  300. case "failed":
  301. taskInfo.Status = model.TaskStatusFailure
  302. default:
  303. return nil, fmt.Errorf("unknown task status: %s", status)
  304. }
  305. if videos := resPayload.Data.TaskResult.Videos; len(videos) > 0 {
  306. video := videos[0]
  307. taskInfo.Url = video.Url
  308. }
  309. return taskInfo, nil
  310. }
  311. func isNewAPIRelay(apiKey string) bool {
  312. return strings.HasPrefix(apiKey, "sk-")
  313. }
  314. func (a *TaskAdaptor) ConvertToOpenAIVideo(originTask *model.Task) ([]byte, error) {
  315. var klingResp responsePayload
  316. if err := common.Unmarshal(originTask.Data, &klingResp); err != nil {
  317. return nil, errors.Wrap(err, "unmarshal kling task data failed")
  318. }
  319. openAIVideo := dto.NewOpenAIVideo()
  320. openAIVideo.ID = originTask.TaskID
  321. openAIVideo.Status = originTask.Status.ToVideoStatus()
  322. openAIVideo.SetProgressStr(originTask.Progress)
  323. openAIVideo.CreatedAt = klingResp.Data.CreatedAt
  324. openAIVideo.CompletedAt = klingResp.Data.UpdatedAt
  325. if len(klingResp.Data.TaskResult.Videos) > 0 {
  326. video := klingResp.Data.TaskResult.Videos[0]
  327. if video.Url != "" {
  328. openAIVideo.SetMetadata("url", video.Url)
  329. }
  330. if video.Duration != "" {
  331. openAIVideo.Seconds = video.Duration
  332. }
  333. }
  334. if klingResp.Code != 0 && klingResp.Message != "" {
  335. openAIVideo.Error = &dto.OpenAIVideoError{
  336. Message: klingResp.Message,
  337. Code: fmt.Sprintf("%d", klingResp.Code),
  338. }
  339. }
  340. return common.Marshal(openAIVideo)
  341. }