adaptor.go 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475
  1. package jimeng
  2. import (
  3. "bytes"
  4. "crypto/hmac"
  5. "crypto/sha256"
  6. "encoding/base64"
  7. "encoding/hex"
  8. "encoding/json"
  9. "fmt"
  10. "io"
  11. "net/http"
  12. "net/url"
  13. "sort"
  14. "strings"
  15. "time"
  16. "github.com/QuantumNous/new-api/model"
  17. "github.com/gin-gonic/gin"
  18. "github.com/pkg/errors"
  19. "github.com/QuantumNous/new-api/constant"
  20. "github.com/QuantumNous/new-api/dto"
  21. "github.com/QuantumNous/new-api/relay/channel"
  22. relaycommon "github.com/QuantumNous/new-api/relay/common"
  23. "github.com/QuantumNous/new-api/service"
  24. )
  25. // ============================
  26. // Request / Response structures
  27. // ============================
  28. type requestPayload struct {
  29. ReqKey string `json:"req_key"`
  30. BinaryDataBase64 []string `json:"binary_data_base64,omitempty"`
  31. ImageUrls []string `json:"image_urls,omitempty"`
  32. Prompt string `json:"prompt,omitempty"`
  33. Seed int64 `json:"seed"`
  34. AspectRatio string `json:"aspect_ratio"`
  35. Frames int `json:"frames,omitempty"`
  36. }
  37. type responsePayload struct {
  38. Code int `json:"code"`
  39. Message string `json:"message"`
  40. RequestId string `json:"request_id"`
  41. Data struct {
  42. TaskID string `json:"task_id"`
  43. } `json:"data"`
  44. }
  45. type responseTask struct {
  46. Code int `json:"code"`
  47. Data struct {
  48. BinaryDataBase64 []interface{} `json:"binary_data_base64"`
  49. ImageUrls interface{} `json:"image_urls"`
  50. RespData string `json:"resp_data"`
  51. Status string `json:"status"`
  52. VideoUrl string `json:"video_url"`
  53. } `json:"data"`
  54. Message string `json:"message"`
  55. RequestId string `json:"request_id"`
  56. Status int `json:"status"`
  57. TimeElapsed string `json:"time_elapsed"`
  58. }
  59. const (
  60. // 即梦限制单个文件最大4.7MB https://www.volcengine.com/docs/85621/1747301
  61. MaxFileSize int64 = 4*1024*1024 + 700*1024 // 4.7MB (4MB + 724KB)
  62. )
  63. // ============================
  64. // Adaptor implementation
  65. // ============================
  66. type TaskAdaptor struct {
  67. ChannelType int
  68. accessKey string
  69. secretKey string
  70. baseURL string
  71. }
  72. func (a *TaskAdaptor) Init(info *relaycommon.RelayInfo) {
  73. a.ChannelType = info.ChannelType
  74. a.baseURL = info.ChannelBaseUrl
  75. // apiKey format: "access_key|secret_key"
  76. keyParts := strings.Split(info.ApiKey, "|")
  77. if len(keyParts) == 2 {
  78. a.accessKey = strings.TrimSpace(keyParts[0])
  79. a.secretKey = strings.TrimSpace(keyParts[1])
  80. }
  81. }
  82. // ValidateRequestAndSetAction parses body, validates fields and sets default action.
  83. func (a *TaskAdaptor) ValidateRequestAndSetAction(c *gin.Context, info *relaycommon.RelayInfo) (taskErr *dto.TaskError) {
  84. return relaycommon.ValidateBasicTaskRequest(c, info, constant.TaskActionGenerate)
  85. }
  86. // BuildRequestURL constructs the upstream URL.
  87. func (a *TaskAdaptor) BuildRequestURL(info *relaycommon.RelayInfo) (string, error) {
  88. if isNewAPIRelay(info.ApiKey) {
  89. return fmt.Sprintf("%s/jimeng/?Action=CVSync2AsyncSubmitTask&Version=2022-08-31", a.baseURL), nil
  90. }
  91. return fmt.Sprintf("%s/?Action=CVSync2AsyncSubmitTask&Version=2022-08-31", a.baseURL), nil
  92. }
  93. // BuildRequestHeader sets required headers.
  94. func (a *TaskAdaptor) BuildRequestHeader(c *gin.Context, req *http.Request, info *relaycommon.RelayInfo) error {
  95. req.Header.Set("Content-Type", "application/json")
  96. req.Header.Set("Accept", "application/json")
  97. if isNewAPIRelay(info.ApiKey) {
  98. req.Header.Set("Authorization", "Bearer "+info.ApiKey)
  99. } else {
  100. return a.signRequest(req, a.accessKey, a.secretKey)
  101. }
  102. return nil
  103. }
  104. func (a *TaskAdaptor) BuildRequestBody(c *gin.Context, info *relaycommon.RelayInfo) (io.Reader, error) {
  105. v, exists := c.Get("task_request")
  106. if !exists {
  107. return nil, fmt.Errorf("request not found in context")
  108. }
  109. req, ok := v.(relaycommon.TaskSubmitReq)
  110. if !ok {
  111. return nil, fmt.Errorf("invalid request type in context")
  112. }
  113. // 支持openai sdk的图片上传方式
  114. if mf, err := c.MultipartForm(); err == nil {
  115. if files, exists := mf.File["input_reference"]; exists && len(files) > 0 {
  116. if len(files) == 1 {
  117. info.Action = constant.TaskActionGenerate
  118. } else if len(files) > 1 {
  119. info.Action = constant.TaskActionFirstTailGenerate
  120. }
  121. // 将上传的文件转换为base64格式
  122. var images []string
  123. for _, fileHeader := range files {
  124. // 检查文件大小
  125. if fileHeader.Size > MaxFileSize {
  126. return nil, fmt.Errorf("文件 %s 大小超过限制,最大允许 %d MB", fileHeader.Filename, MaxFileSize/(1024*1024))
  127. }
  128. file, err := fileHeader.Open()
  129. if err != nil {
  130. continue
  131. }
  132. fileBytes, err := io.ReadAll(file)
  133. file.Close()
  134. if err != nil {
  135. continue
  136. }
  137. // 将文件内容转换为base64
  138. base64Str := base64.StdEncoding.EncodeToString(fileBytes)
  139. images = append(images, base64Str)
  140. }
  141. req.Images = images
  142. }
  143. }
  144. body, err := a.convertToRequestPayload(&req)
  145. if err != nil {
  146. return nil, errors.Wrap(err, "convert request payload failed")
  147. }
  148. data, err := json.Marshal(body)
  149. if err != nil {
  150. return nil, err
  151. }
  152. return bytes.NewReader(data), nil
  153. }
  154. // DoRequest delegates to common helper.
  155. func (a *TaskAdaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
  156. return channel.DoTaskApiRequest(a, c, info, requestBody)
  157. }
  158. // DoResponse handles upstream response, returns taskID etc.
  159. func (a *TaskAdaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (taskID string, taskData []byte, taskErr *dto.TaskError) {
  160. responseBody, err := io.ReadAll(resp.Body)
  161. if err != nil {
  162. taskErr = service.TaskErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError)
  163. return
  164. }
  165. _ = resp.Body.Close()
  166. // Parse Jimeng response
  167. var jResp responsePayload
  168. if err := json.Unmarshal(responseBody, &jResp); err != nil {
  169. taskErr = service.TaskErrorWrapper(errors.Wrapf(err, "body: %s", responseBody), "unmarshal_response_body_failed", http.StatusInternalServerError)
  170. return
  171. }
  172. if jResp.Code != 10000 {
  173. taskErr = service.TaskErrorWrapper(fmt.Errorf(jResp.Message), fmt.Sprintf("%d", jResp.Code), http.StatusInternalServerError)
  174. return
  175. }
  176. ov := dto.NewOpenAIVideo()
  177. ov.ID = jResp.Data.TaskID
  178. ov.TaskID = jResp.Data.TaskID
  179. ov.CreatedAt = time.Now().Unix()
  180. ov.Model = info.OriginModelName
  181. c.JSON(http.StatusOK, ov)
  182. return jResp.Data.TaskID, responseBody, nil
  183. }
  184. // FetchTask fetch task status
  185. func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any) (*http.Response, error) {
  186. taskID, ok := body["task_id"].(string)
  187. if !ok {
  188. return nil, fmt.Errorf("invalid task_id")
  189. }
  190. uri := fmt.Sprintf("%s/?Action=CVSync2AsyncGetResult&Version=2022-08-31", baseUrl)
  191. if isNewAPIRelay(key) {
  192. uri = fmt.Sprintf("%s/jimeng/?Action=CVSync2AsyncGetResult&Version=2022-08-31", a.baseURL)
  193. }
  194. payload := map[string]string{
  195. "req_key": "jimeng_vgfm_t2v_l20", // This is fixed value from doc: https://www.volcengine.com/docs/85621/1544774
  196. "task_id": taskID,
  197. }
  198. payloadBytes, err := json.Marshal(payload)
  199. if err != nil {
  200. return nil, errors.Wrap(err, "marshal fetch task payload failed")
  201. }
  202. req, err := http.NewRequest(http.MethodPost, uri, bytes.NewBuffer(payloadBytes))
  203. if err != nil {
  204. return nil, err
  205. }
  206. req.Header.Set("Accept", "application/json")
  207. req.Header.Set("Content-Type", "application/json")
  208. if isNewAPIRelay(key) {
  209. req.Header.Set("Authorization", "Bearer "+key)
  210. } else {
  211. keyParts := strings.Split(key, "|")
  212. if len(keyParts) != 2 {
  213. return nil, fmt.Errorf("invalid api key format for jimeng: expected 'ak|sk'")
  214. }
  215. accessKey := strings.TrimSpace(keyParts[0])
  216. secretKey := strings.TrimSpace(keyParts[1])
  217. if err := a.signRequest(req, accessKey, secretKey); err != nil {
  218. return nil, errors.Wrap(err, "sign request failed")
  219. }
  220. }
  221. return service.GetHttpClient().Do(req)
  222. }
  223. func (a *TaskAdaptor) GetModelList() []string {
  224. return []string{"jimeng_vgfm_t2v_l20"}
  225. }
  226. func (a *TaskAdaptor) GetChannelName() string {
  227. return "jimeng"
  228. }
  229. func (a *TaskAdaptor) signRequest(req *http.Request, accessKey, secretKey string) error {
  230. var bodyBytes []byte
  231. var err error
  232. if req.Body != nil {
  233. bodyBytes, err = io.ReadAll(req.Body)
  234. if err != nil {
  235. return errors.Wrap(err, "read request body failed")
  236. }
  237. _ = req.Body.Close()
  238. req.Body = io.NopCloser(bytes.NewBuffer(bodyBytes)) // Rewind
  239. } else {
  240. bodyBytes = []byte{}
  241. }
  242. payloadHash := sha256.Sum256(bodyBytes)
  243. hexPayloadHash := hex.EncodeToString(payloadHash[:])
  244. t := time.Now().UTC()
  245. xDate := t.Format("20060102T150405Z")
  246. shortDate := t.Format("20060102")
  247. req.Header.Set("Host", req.URL.Host)
  248. req.Header.Set("X-Date", xDate)
  249. req.Header.Set("X-Content-Sha256", hexPayloadHash)
  250. // Sort and encode query parameters to create canonical query string
  251. queryParams := req.URL.Query()
  252. sortedKeys := make([]string, 0, len(queryParams))
  253. for k := range queryParams {
  254. sortedKeys = append(sortedKeys, k)
  255. }
  256. sort.Strings(sortedKeys)
  257. var queryParts []string
  258. for _, k := range sortedKeys {
  259. values := queryParams[k]
  260. sort.Strings(values)
  261. for _, v := range values {
  262. queryParts = append(queryParts, fmt.Sprintf("%s=%s", url.QueryEscape(k), url.QueryEscape(v)))
  263. }
  264. }
  265. canonicalQueryString := strings.Join(queryParts, "&")
  266. headersToSign := map[string]string{
  267. "host": req.URL.Host,
  268. "x-date": xDate,
  269. "x-content-sha256": hexPayloadHash,
  270. }
  271. if req.Header.Get("Content-Type") != "" {
  272. headersToSign["content-type"] = req.Header.Get("Content-Type")
  273. }
  274. var signedHeaderKeys []string
  275. for k := range headersToSign {
  276. signedHeaderKeys = append(signedHeaderKeys, k)
  277. }
  278. sort.Strings(signedHeaderKeys)
  279. var canonicalHeaders strings.Builder
  280. for _, k := range signedHeaderKeys {
  281. canonicalHeaders.WriteString(k)
  282. canonicalHeaders.WriteString(":")
  283. canonicalHeaders.WriteString(strings.TrimSpace(headersToSign[k]))
  284. canonicalHeaders.WriteString("\n")
  285. }
  286. signedHeaders := strings.Join(signedHeaderKeys, ";")
  287. canonicalRequest := fmt.Sprintf("%s\n%s\n%s\n%s\n%s\n%s",
  288. req.Method,
  289. req.URL.Path,
  290. canonicalQueryString,
  291. canonicalHeaders.String(),
  292. signedHeaders,
  293. hexPayloadHash,
  294. )
  295. hashedCanonicalRequest := sha256.Sum256([]byte(canonicalRequest))
  296. hexHashedCanonicalRequest := hex.EncodeToString(hashedCanonicalRequest[:])
  297. region := "cn-north-1"
  298. serviceName := "cv"
  299. credentialScope := fmt.Sprintf("%s/%s/%s/request", shortDate, region, serviceName)
  300. stringToSign := fmt.Sprintf("HMAC-SHA256\n%s\n%s\n%s",
  301. xDate,
  302. credentialScope,
  303. hexHashedCanonicalRequest,
  304. )
  305. kDate := hmacSHA256([]byte(secretKey), []byte(shortDate))
  306. kRegion := hmacSHA256(kDate, []byte(region))
  307. kService := hmacSHA256(kRegion, []byte(serviceName))
  308. kSigning := hmacSHA256(kService, []byte("request"))
  309. signature := hex.EncodeToString(hmacSHA256(kSigning, []byte(stringToSign)))
  310. authorization := fmt.Sprintf("HMAC-SHA256 Credential=%s/%s, SignedHeaders=%s, Signature=%s",
  311. accessKey,
  312. credentialScope,
  313. signedHeaders,
  314. signature,
  315. )
  316. req.Header.Set("Authorization", authorization)
  317. return nil
  318. }
  319. func hmacSHA256(key []byte, data []byte) []byte {
  320. h := hmac.New(sha256.New, key)
  321. h.Write(data)
  322. return h.Sum(nil)
  323. }
  324. func (a *TaskAdaptor) convertToRequestPayload(req *relaycommon.TaskSubmitReq) (*requestPayload, error) {
  325. r := requestPayload{
  326. ReqKey: req.Model,
  327. Prompt: req.Prompt,
  328. }
  329. switch req.Duration {
  330. case 10:
  331. r.Frames = 241 // 24*10+1 = 241
  332. default:
  333. r.Frames = 121 // 24*5+1 = 121
  334. }
  335. // Handle one-of image_urls or binary_data_base64
  336. if req.HasImage() {
  337. if strings.HasPrefix(req.Images[0], "http") {
  338. r.ImageUrls = req.Images
  339. } else {
  340. r.BinaryDataBase64 = req.Images
  341. }
  342. }
  343. metadata := req.Metadata
  344. medaBytes, err := json.Marshal(metadata)
  345. if err != nil {
  346. return nil, errors.Wrap(err, "metadata marshal metadata failed")
  347. }
  348. err = json.Unmarshal(medaBytes, &r)
  349. if err != nil {
  350. return nil, errors.Wrap(err, "unmarshal metadata failed")
  351. }
  352. // 即梦视频3.0 ReqKey转换
  353. // https://www.volcengine.com/docs/85621/1792707
  354. if strings.Contains(r.ReqKey, "jimeng_v30") {
  355. if len(req.Images) > 1 {
  356. // 多张图片:首尾帧生成
  357. r.ReqKey = strings.Replace(r.ReqKey, "jimeng_v30", "jimeng_i2v_first_tail_v30", 1)
  358. } else if len(req.Images) == 1 {
  359. // 单张图片:图生视频
  360. r.ReqKey = strings.Replace(r.ReqKey, "jimeng_v30", "jimeng_i2v_first_v30", 1)
  361. } else {
  362. // 无图片:文生视频
  363. r.ReqKey = strings.Replace(r.ReqKey, "jimeng_v30", "jimeng_t2v_v30", 1)
  364. }
  365. }
  366. return &r, nil
  367. }
  368. func (a *TaskAdaptor) ParseTaskResult(respBody []byte) (*relaycommon.TaskInfo, error) {
  369. resTask := responseTask{}
  370. if err := json.Unmarshal(respBody, &resTask); err != nil {
  371. return nil, errors.Wrap(err, "unmarshal task result failed")
  372. }
  373. taskResult := relaycommon.TaskInfo{}
  374. if resTask.Code == 10000 {
  375. taskResult.Code = 0
  376. } else {
  377. taskResult.Code = resTask.Code // todo uni code
  378. taskResult.Reason = resTask.Message
  379. taskResult.Status = model.TaskStatusFailure
  380. taskResult.Progress = "100%"
  381. }
  382. switch resTask.Data.Status {
  383. case "in_queue":
  384. taskResult.Status = model.TaskStatusQueued
  385. taskResult.Progress = "10%"
  386. case "done":
  387. taskResult.Status = model.TaskStatusSuccess
  388. taskResult.Progress = "100%"
  389. }
  390. taskResult.Url = resTask.Data.VideoUrl
  391. return &taskResult, nil
  392. }
  393. func (a *TaskAdaptor) ConvertToOpenAIVideo(originTask *model.Task) (*dto.OpenAIVideo, error) {
  394. var jimengResp responseTask
  395. if err := json.Unmarshal(originTask.Data, &jimengResp); err != nil {
  396. return nil, errors.Wrap(err, "unmarshal jimeng task data failed")
  397. }
  398. openAIVideo := dto.NewOpenAIVideo()
  399. openAIVideo.ID = originTask.TaskID
  400. openAIVideo.Status = originTask.Status.ToVideoStatus()
  401. openAIVideo.SetProgressStr(originTask.Progress)
  402. openAIVideo.SetMetadata("url", jimengResp.Data.VideoUrl)
  403. openAIVideo.CreatedAt = originTask.CreatedAt
  404. openAIVideo.CompletedAt = originTask.UpdatedAt
  405. if jimengResp.Code != 10000 {
  406. openAIVideo.Error = &dto.OpenAIVideoError{
  407. Message: jimengResp.Message,
  408. Code: fmt.Sprintf("%d", jimengResp.Code),
  409. }
  410. }
  411. return openAIVideo, nil
  412. }
  413. func isNewAPIRelay(apiKey string) bool {
  414. return strings.HasPrefix(apiKey, "sk-")
  415. }