adaptor.go 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484
  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/common"
  17. "github.com/QuantumNous/new-api/model"
  18. "github.com/gin-gonic/gin"
  19. "github.com/pkg/errors"
  20. "github.com/QuantumNous/new-api/constant"
  21. "github.com/QuantumNous/new-api/dto"
  22. "github.com/QuantumNous/new-api/relay/channel"
  23. relaycommon "github.com/QuantumNous/new-api/relay/common"
  24. "github.com/QuantumNous/new-api/service"
  25. )
  26. // ============================
  27. // Request / Response structures
  28. // ============================
  29. type requestPayload struct {
  30. ReqKey string `json:"req_key"`
  31. BinaryDataBase64 []string `json:"binary_data_base64,omitempty"`
  32. ImageUrls []string `json:"image_urls,omitempty"`
  33. Prompt string `json:"prompt,omitempty"`
  34. Seed int64 `json:"seed"`
  35. AspectRatio string `json:"aspect_ratio"`
  36. Frames int `json:"frames,omitempty"`
  37. }
  38. type responsePayload struct {
  39. Code int `json:"code"`
  40. Message string `json:"message"`
  41. RequestId string `json:"request_id"`
  42. Data struct {
  43. TaskID string `json:"task_id"`
  44. } `json:"data"`
  45. }
  46. type responseTask struct {
  47. Code int `json:"code"`
  48. Data struct {
  49. BinaryDataBase64 []interface{} `json:"binary_data_base64"`
  50. ImageUrls interface{} `json:"image_urls"`
  51. RespData string `json:"resp_data"`
  52. Status string `json:"status"`
  53. VideoUrl string `json:"video_url"`
  54. } `json:"data"`
  55. Message string `json:"message"`
  56. RequestId string `json:"request_id"`
  57. Status int `json:"status"`
  58. TimeElapsed string `json:"time_elapsed"`
  59. }
  60. const (
  61. // 即梦限制单个文件最大4.7MB https://www.volcengine.com/docs/85621/1747301
  62. MaxFileSize int64 = 4*1024*1024 + 700*1024 // 4.7MB (4MB + 724KB)
  63. )
  64. // ============================
  65. // Adaptor implementation
  66. // ============================
  67. type TaskAdaptor struct {
  68. ChannelType int
  69. accessKey string
  70. secretKey string
  71. baseURL string
  72. }
  73. func (a *TaskAdaptor) Init(info *relaycommon.RelayInfo) {
  74. a.ChannelType = info.ChannelType
  75. a.baseURL = info.ChannelBaseUrl
  76. // apiKey format: "access_key|secret_key"
  77. keyParts := strings.Split(info.ApiKey, "|")
  78. if len(keyParts) == 2 {
  79. a.accessKey = strings.TrimSpace(keyParts[0])
  80. a.secretKey = strings.TrimSpace(keyParts[1])
  81. }
  82. }
  83. // ValidateRequestAndSetAction parses body, validates fields and sets default action.
  84. func (a *TaskAdaptor) ValidateRequestAndSetAction(c *gin.Context, info *relaycommon.RelayInfo) (taskErr *dto.TaskError) {
  85. return relaycommon.ValidateBasicTaskRequest(c, info, constant.TaskActionGenerate)
  86. }
  87. // BuildRequestURL constructs the upstream URL.
  88. func (a *TaskAdaptor) BuildRequestURL(info *relaycommon.RelayInfo) (string, error) {
  89. if isNewAPIRelay(info.ApiKey) {
  90. return fmt.Sprintf("%s/jimeng/?Action=CVSync2AsyncSubmitTask&Version=2022-08-31", a.baseURL), nil
  91. }
  92. return fmt.Sprintf("%s/?Action=CVSync2AsyncSubmitTask&Version=2022-08-31", a.baseURL), nil
  93. }
  94. // BuildRequestHeader sets required headers.
  95. func (a *TaskAdaptor) BuildRequestHeader(c *gin.Context, req *http.Request, info *relaycommon.RelayInfo) error {
  96. req.Header.Set("Content-Type", "application/json")
  97. req.Header.Set("Accept", "application/json")
  98. if isNewAPIRelay(info.ApiKey) {
  99. req.Header.Set("Authorization", "Bearer "+info.ApiKey)
  100. } else {
  101. return a.signRequest(req, a.accessKey, a.secretKey)
  102. }
  103. return nil
  104. }
  105. func (a *TaskAdaptor) BuildRequestBody(c *gin.Context, info *relaycommon.RelayInfo) (io.Reader, error) {
  106. v, exists := c.Get("task_request")
  107. if !exists {
  108. return nil, fmt.Errorf("request not found in context")
  109. }
  110. req, ok := v.(relaycommon.TaskSubmitReq)
  111. if !ok {
  112. return nil, fmt.Errorf("invalid request type in context")
  113. }
  114. // 支持openai sdk的图片上传方式
  115. if mf, err := c.MultipartForm(); err == nil {
  116. if files, exists := mf.File["input_reference"]; exists && len(files) > 0 {
  117. if len(files) == 1 {
  118. info.Action = constant.TaskActionGenerate
  119. } else if len(files) > 1 {
  120. info.Action = constant.TaskActionFirstTailGenerate
  121. }
  122. // 将上传的文件转换为base64格式
  123. var images []string
  124. for _, fileHeader := range files {
  125. // 检查文件大小
  126. if fileHeader.Size > MaxFileSize {
  127. return nil, fmt.Errorf("文件 %s 大小超过限制,最大允许 %d MB", fileHeader.Filename, MaxFileSize/(1024*1024))
  128. }
  129. file, err := fileHeader.Open()
  130. if err != nil {
  131. continue
  132. }
  133. fileBytes, err := io.ReadAll(file)
  134. file.Close()
  135. if err != nil {
  136. continue
  137. }
  138. // 将文件内容转换为base64
  139. base64Str := base64.StdEncoding.EncodeToString(fileBytes)
  140. images = append(images, base64Str)
  141. }
  142. req.Images = images
  143. }
  144. }
  145. body, err := a.convertToRequestPayload(&req)
  146. if err != nil {
  147. return nil, errors.Wrap(err, "convert request payload failed")
  148. }
  149. data, err := json.Marshal(body)
  150. if err != nil {
  151. return nil, err
  152. }
  153. return bytes.NewReader(data), nil
  154. }
  155. // DoRequest delegates to common helper.
  156. func (a *TaskAdaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (*http.Response, error) {
  157. return channel.DoTaskApiRequest(a, c, info, requestBody)
  158. }
  159. // DoResponse handles upstream response, returns taskID etc.
  160. func (a *TaskAdaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (taskID string, taskData []byte, taskErr *dto.TaskError) {
  161. responseBody, err := io.ReadAll(resp.Body)
  162. if err != nil {
  163. taskErr = service.TaskErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError)
  164. return
  165. }
  166. _ = resp.Body.Close()
  167. // Parse Jimeng response
  168. var jResp responsePayload
  169. if err := json.Unmarshal(responseBody, &jResp); err != nil {
  170. taskErr = service.TaskErrorWrapper(errors.Wrapf(err, "body: %s", responseBody), "unmarshal_response_body_failed", http.StatusInternalServerError)
  171. return
  172. }
  173. if jResp.Code != 10000 {
  174. taskErr = service.TaskErrorWrapper(fmt.Errorf("%s", jResp.Message), fmt.Sprintf("%d", jResp.Code), http.StatusInternalServerError)
  175. return
  176. }
  177. ov := dto.NewOpenAIVideo()
  178. ov.ID = jResp.Data.TaskID
  179. ov.TaskID = jResp.Data.TaskID
  180. ov.CreatedAt = time.Now().Unix()
  181. ov.Model = info.OriginModelName
  182. c.JSON(http.StatusOK, ov)
  183. return jResp.Data.TaskID, responseBody, nil
  184. }
  185. // FetchTask fetch task status
  186. func (a *TaskAdaptor) FetchTask(baseUrl, key string, body map[string]any, proxy string) (*http.Response, error) {
  187. taskID, ok := body["task_id"].(string)
  188. if !ok {
  189. return nil, fmt.Errorf("invalid task_id")
  190. }
  191. uri := fmt.Sprintf("%s/?Action=CVSync2AsyncGetResult&Version=2022-08-31", baseUrl)
  192. if isNewAPIRelay(key) {
  193. uri = fmt.Sprintf("%s/jimeng/?Action=CVSync2AsyncGetResult&Version=2022-08-31", a.baseURL)
  194. }
  195. payload := map[string]string{
  196. "req_key": "jimeng_vgfm_t2v_l20", // This is fixed value from doc: https://www.volcengine.com/docs/85621/1544774
  197. "task_id": taskID,
  198. }
  199. payloadBytes, err := json.Marshal(payload)
  200. if err != nil {
  201. return nil, errors.Wrap(err, "marshal fetch task payload failed")
  202. }
  203. req, err := http.NewRequest(http.MethodPost, uri, bytes.NewBuffer(payloadBytes))
  204. if err != nil {
  205. return nil, err
  206. }
  207. req.Header.Set("Accept", "application/json")
  208. req.Header.Set("Content-Type", "application/json")
  209. if isNewAPIRelay(key) {
  210. req.Header.Set("Authorization", "Bearer "+key)
  211. } else {
  212. keyParts := strings.Split(key, "|")
  213. if len(keyParts) != 2 {
  214. return nil, fmt.Errorf("invalid api key format for jimeng: expected 'ak|sk'")
  215. }
  216. accessKey := strings.TrimSpace(keyParts[0])
  217. secretKey := strings.TrimSpace(keyParts[1])
  218. if err := a.signRequest(req, accessKey, secretKey); err != nil {
  219. return nil, errors.Wrap(err, "sign request failed")
  220. }
  221. }
  222. client, err := service.GetHttpClientWithProxy(proxy)
  223. if err != nil {
  224. return nil, fmt.Errorf("new proxy http client failed: %w", err)
  225. }
  226. return client.Do(req)
  227. }
  228. func (a *TaskAdaptor) GetModelList() []string {
  229. return []string{"jimeng_vgfm_t2v_l20"}
  230. }
  231. func (a *TaskAdaptor) GetChannelName() string {
  232. return "jimeng"
  233. }
  234. func (a *TaskAdaptor) signRequest(req *http.Request, accessKey, secretKey string) error {
  235. var bodyBytes []byte
  236. var err error
  237. if req.Body != nil {
  238. bodyBytes, err = io.ReadAll(req.Body)
  239. if err != nil {
  240. return errors.Wrap(err, "read request body failed")
  241. }
  242. _ = req.Body.Close()
  243. req.Body = io.NopCloser(bytes.NewBuffer(bodyBytes)) // Rewind
  244. } else {
  245. bodyBytes = []byte{}
  246. }
  247. payloadHash := sha256.Sum256(bodyBytes)
  248. hexPayloadHash := hex.EncodeToString(payloadHash[:])
  249. t := time.Now().UTC()
  250. xDate := t.Format("20060102T150405Z")
  251. shortDate := t.Format("20060102")
  252. req.Header.Set("Host", req.URL.Host)
  253. req.Header.Set("X-Date", xDate)
  254. req.Header.Set("X-Content-Sha256", hexPayloadHash)
  255. // Sort and encode query parameters to create canonical query string
  256. queryParams := req.URL.Query()
  257. sortedKeys := make([]string, 0, len(queryParams))
  258. for k := range queryParams {
  259. sortedKeys = append(sortedKeys, k)
  260. }
  261. sort.Strings(sortedKeys)
  262. var queryParts []string
  263. for _, k := range sortedKeys {
  264. values := queryParams[k]
  265. sort.Strings(values)
  266. for _, v := range values {
  267. queryParts = append(queryParts, fmt.Sprintf("%s=%s", url.QueryEscape(k), url.QueryEscape(v)))
  268. }
  269. }
  270. canonicalQueryString := strings.Join(queryParts, "&")
  271. headersToSign := map[string]string{
  272. "host": req.URL.Host,
  273. "x-date": xDate,
  274. "x-content-sha256": hexPayloadHash,
  275. }
  276. if req.Header.Get("Content-Type") != "" {
  277. headersToSign["content-type"] = req.Header.Get("Content-Type")
  278. }
  279. var signedHeaderKeys []string
  280. for k := range headersToSign {
  281. signedHeaderKeys = append(signedHeaderKeys, k)
  282. }
  283. sort.Strings(signedHeaderKeys)
  284. var canonicalHeaders strings.Builder
  285. for _, k := range signedHeaderKeys {
  286. canonicalHeaders.WriteString(k)
  287. canonicalHeaders.WriteString(":")
  288. canonicalHeaders.WriteString(strings.TrimSpace(headersToSign[k]))
  289. canonicalHeaders.WriteString("\n")
  290. }
  291. signedHeaders := strings.Join(signedHeaderKeys, ";")
  292. canonicalRequest := fmt.Sprintf("%s\n%s\n%s\n%s\n%s\n%s",
  293. req.Method,
  294. req.URL.Path,
  295. canonicalQueryString,
  296. canonicalHeaders.String(),
  297. signedHeaders,
  298. hexPayloadHash,
  299. )
  300. hashedCanonicalRequest := sha256.Sum256([]byte(canonicalRequest))
  301. hexHashedCanonicalRequest := hex.EncodeToString(hashedCanonicalRequest[:])
  302. region := "cn-north-1"
  303. serviceName := "cv"
  304. credentialScope := fmt.Sprintf("%s/%s/%s/request", shortDate, region, serviceName)
  305. stringToSign := fmt.Sprintf("HMAC-SHA256\n%s\n%s\n%s",
  306. xDate,
  307. credentialScope,
  308. hexHashedCanonicalRequest,
  309. )
  310. kDate := hmacSHA256([]byte(secretKey), []byte(shortDate))
  311. kRegion := hmacSHA256(kDate, []byte(region))
  312. kService := hmacSHA256(kRegion, []byte(serviceName))
  313. kSigning := hmacSHA256(kService, []byte("request"))
  314. signature := hex.EncodeToString(hmacSHA256(kSigning, []byte(stringToSign)))
  315. authorization := fmt.Sprintf("HMAC-SHA256 Credential=%s/%s, SignedHeaders=%s, Signature=%s",
  316. accessKey,
  317. credentialScope,
  318. signedHeaders,
  319. signature,
  320. )
  321. req.Header.Set("Authorization", authorization)
  322. return nil
  323. }
  324. func hmacSHA256(key []byte, data []byte) []byte {
  325. h := hmac.New(sha256.New, key)
  326. h.Write(data)
  327. return h.Sum(nil)
  328. }
  329. func (a *TaskAdaptor) convertToRequestPayload(req *relaycommon.TaskSubmitReq) (*requestPayload, error) {
  330. r := requestPayload{
  331. ReqKey: req.Model,
  332. Prompt: req.Prompt,
  333. }
  334. switch req.Duration {
  335. case 10:
  336. r.Frames = 241 // 24*10+1 = 241
  337. default:
  338. r.Frames = 121 // 24*5+1 = 121
  339. }
  340. // Handle one-of image_urls or binary_data_base64
  341. if req.HasImage() {
  342. if strings.HasPrefix(req.Images[0], "http") {
  343. r.ImageUrls = req.Images
  344. } else {
  345. r.BinaryDataBase64 = req.Images
  346. }
  347. }
  348. metadata := req.Metadata
  349. medaBytes, err := json.Marshal(metadata)
  350. if err != nil {
  351. return nil, errors.Wrap(err, "metadata marshal metadata failed")
  352. }
  353. err = json.Unmarshal(medaBytes, &r)
  354. if err != nil {
  355. return nil, errors.Wrap(err, "unmarshal metadata failed")
  356. }
  357. // 即梦视频3.0 ReqKey转换
  358. // https://www.volcengine.com/docs/85621/1792707
  359. if strings.Contains(r.ReqKey, "jimeng_v30") {
  360. if r.ReqKey == "jimeng_v30_pro" {
  361. // 3.0 pro只有固定的jimeng_ti2v_v30_pro
  362. r.ReqKey = "jimeng_ti2v_v30_pro"
  363. } else if len(req.Images) > 1 {
  364. // 多张图片:首尾帧生成
  365. r.ReqKey = strings.TrimSuffix(strings.Replace(r.ReqKey, "jimeng_v30", "jimeng_i2v_first_tail_v30", 1), "p")
  366. } else if len(req.Images) == 1 {
  367. // 单张图片:图生视频
  368. r.ReqKey = strings.TrimSuffix(strings.Replace(r.ReqKey, "jimeng_v30", "jimeng_i2v_first_v30", 1), "p")
  369. } else {
  370. // 无图片:文生视频
  371. r.ReqKey = strings.Replace(r.ReqKey, "jimeng_v30", "jimeng_t2v_v30", 1)
  372. }
  373. }
  374. return &r, nil
  375. }
  376. func (a *TaskAdaptor) ParseTaskResult(respBody []byte) (*relaycommon.TaskInfo, error) {
  377. resTask := responseTask{}
  378. if err := json.Unmarshal(respBody, &resTask); err != nil {
  379. return nil, errors.Wrap(err, "unmarshal task result failed")
  380. }
  381. taskResult := relaycommon.TaskInfo{}
  382. if resTask.Code == 10000 {
  383. taskResult.Code = 0
  384. } else {
  385. taskResult.Code = resTask.Code // todo uni code
  386. taskResult.Reason = resTask.Message
  387. taskResult.Status = model.TaskStatusFailure
  388. taskResult.Progress = "100%"
  389. }
  390. switch resTask.Data.Status {
  391. case "in_queue":
  392. taskResult.Status = model.TaskStatusQueued
  393. taskResult.Progress = "10%"
  394. case "done":
  395. taskResult.Status = model.TaskStatusSuccess
  396. taskResult.Progress = "100%"
  397. }
  398. taskResult.Url = resTask.Data.VideoUrl
  399. return &taskResult, nil
  400. }
  401. func (a *TaskAdaptor) ConvertToOpenAIVideo(originTask *model.Task) ([]byte, error) {
  402. var jimengResp responseTask
  403. if err := json.Unmarshal(originTask.Data, &jimengResp); err != nil {
  404. return nil, errors.Wrap(err, "unmarshal jimeng task data failed")
  405. }
  406. openAIVideo := dto.NewOpenAIVideo()
  407. openAIVideo.ID = originTask.TaskID
  408. openAIVideo.Status = originTask.Status.ToVideoStatus()
  409. openAIVideo.SetProgressStr(originTask.Progress)
  410. openAIVideo.SetMetadata("url", jimengResp.Data.VideoUrl)
  411. openAIVideo.CreatedAt = originTask.CreatedAt
  412. openAIVideo.CompletedAt = originTask.UpdatedAt
  413. if jimengResp.Code != 10000 {
  414. openAIVideo.Error = &dto.OpenAIVideoError{
  415. Message: jimengResp.Message,
  416. Code: fmt.Sprintf("%d", jimengResp.Code),
  417. }
  418. }
  419. jsonData, _ := common.Marshal(openAIVideo)
  420. return jsonData, nil
  421. }
  422. func isNewAPIRelay(apiKey string) bool {
  423. return strings.HasPrefix(apiKey, "sk-")
  424. }