relay-dify.go 8.7 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297
  1. package dify
  2. import (
  3. "bytes"
  4. "encoding/base64"
  5. "encoding/json"
  6. "fmt"
  7. "io"
  8. "mime/multipart"
  9. "net/http"
  10. "os"
  11. "strings"
  12. "github.com/QuantumNous/new-api/common"
  13. "github.com/QuantumNous/new-api/constant"
  14. "github.com/QuantumNous/new-api/dto"
  15. relaycommon "github.com/QuantumNous/new-api/relay/common"
  16. "github.com/QuantumNous/new-api/relay/helper"
  17. "github.com/QuantumNous/new-api/service"
  18. "github.com/QuantumNous/new-api/types"
  19. "github.com/samber/lo"
  20. "github.com/gin-gonic/gin"
  21. )
  22. func uploadDifyFile(c *gin.Context, info *relaycommon.RelayInfo, user string, media dto.MediaContent) *DifyFile {
  23. uploadUrl := fmt.Sprintf("%s/v1/files/upload", info.ChannelBaseUrl)
  24. switch media.Type {
  25. case dto.ContentTypeImageURL:
  26. // Decode base64 data
  27. imageMedia := media.GetImageMedia()
  28. base64Data := imageMedia.Url
  29. // Remove base64 prefix if exists (e.g., "data:image/jpeg;base64,")
  30. if idx := strings.Index(base64Data, ","); idx != -1 {
  31. base64Data = base64Data[idx+1:]
  32. }
  33. // Decode base64 string
  34. decodedData, err := base64.StdEncoding.DecodeString(base64Data)
  35. if err != nil {
  36. common.SysLog("failed to decode base64: " + err.Error())
  37. return nil
  38. }
  39. // Create temporary file
  40. tempFile, err := os.CreateTemp("", "dify-upload-*")
  41. if err != nil {
  42. common.SysLog("failed to create temp file: " + err.Error())
  43. return nil
  44. }
  45. defer tempFile.Close()
  46. defer os.Remove(tempFile.Name())
  47. // Write decoded data to temp file
  48. if _, err := tempFile.Write(decodedData); err != nil {
  49. common.SysLog("failed to write to temp file: " + err.Error())
  50. return nil
  51. }
  52. // Create multipart form
  53. body := &bytes.Buffer{}
  54. writer := multipart.NewWriter(body)
  55. // Add user field
  56. if err := writer.WriteField("user", user); err != nil {
  57. common.SysLog("failed to add user field: " + err.Error())
  58. return nil
  59. }
  60. // Create form file with proper mime type
  61. mimeType := imageMedia.MimeType
  62. if mimeType == "" {
  63. mimeType = "image/jpeg" // default mime type
  64. }
  65. // Create form file
  66. part, err := writer.CreateFormFile("file", fmt.Sprintf("image.%s", strings.TrimPrefix(mimeType, "image/")))
  67. if err != nil {
  68. common.SysLog("failed to create form file: " + err.Error())
  69. return nil
  70. }
  71. // Copy file content to form
  72. if _, err = io.Copy(part, bytes.NewReader(decodedData)); err != nil {
  73. common.SysLog("failed to copy file content: " + err.Error())
  74. return nil
  75. }
  76. writer.Close()
  77. // Create HTTP request
  78. req, err := http.NewRequest("POST", uploadUrl, body)
  79. if err != nil {
  80. common.SysLog("failed to create request: " + err.Error())
  81. return nil
  82. }
  83. req.Header.Set("Content-Type", writer.FormDataContentType())
  84. req.Header.Set("Authorization", fmt.Sprintf("Bearer %s", info.ApiKey))
  85. // Send request
  86. client := service.GetHttpClient()
  87. resp, err := client.Do(req)
  88. if err != nil {
  89. common.SysLog("failed to send request: " + err.Error())
  90. return nil
  91. }
  92. defer resp.Body.Close()
  93. // Parse response
  94. var result struct {
  95. Id string `json:"id"`
  96. }
  97. if err := json.NewDecoder(resp.Body).Decode(&result); err != nil {
  98. common.SysLog("failed to decode response: " + err.Error())
  99. return nil
  100. }
  101. return &DifyFile{
  102. UploadFileId: result.Id,
  103. Type: "image",
  104. TransferMode: "local_file",
  105. }
  106. }
  107. return nil
  108. }
  109. func requestOpenAI2Dify(c *gin.Context, info *relaycommon.RelayInfo, request dto.GeneralOpenAIRequest) *DifyChatRequest {
  110. difyReq := DifyChatRequest{
  111. Inputs: make(map[string]interface{}),
  112. AutoGenerateName: false,
  113. }
  114. user := request.User
  115. if len(user) == 0 {
  116. user = json.RawMessage(helper.GetResponseID(c))
  117. }
  118. var stringUser string
  119. err := json.Unmarshal(user, &stringUser)
  120. if err != nil {
  121. common.SysLog("failed to unmarshal user: " + err.Error())
  122. stringUser = helper.GetResponseID(c)
  123. }
  124. difyReq.User = stringUser
  125. files := make([]DifyFile, 0)
  126. var content strings.Builder
  127. for _, message := range request.Messages {
  128. if message.Role == "system" {
  129. content.WriteString("SYSTEM: \n" + message.StringContent() + "\n")
  130. } else if message.Role == "assistant" {
  131. content.WriteString("ASSISTANT: \n" + message.StringContent() + "\n")
  132. } else {
  133. parseContent := message.ParseContent()
  134. for _, mediaContent := range parseContent {
  135. switch mediaContent.Type {
  136. case dto.ContentTypeText:
  137. content.WriteString("USER: \n" + mediaContent.Text + "\n")
  138. case dto.ContentTypeImageURL:
  139. media := mediaContent.GetImageMedia()
  140. var file *DifyFile
  141. if media.IsRemoteImage() {
  142. file.Type = media.MimeType
  143. file.TransferMode = "remote_url"
  144. file.URL = media.Url
  145. } else {
  146. file = uploadDifyFile(c, info, difyReq.User, mediaContent)
  147. }
  148. if file != nil {
  149. files = append(files, *file)
  150. }
  151. }
  152. }
  153. }
  154. }
  155. difyReq.Query = content.String()
  156. difyReq.Files = files
  157. mode := "blocking"
  158. if lo.FromPtrOr(request.Stream, false) {
  159. mode = "streaming"
  160. }
  161. difyReq.ResponseMode = mode
  162. return &difyReq
  163. }
  164. func streamResponseDify2OpenAI(difyResponse DifyChunkChatCompletionResponse) *dto.ChatCompletionsStreamResponse {
  165. response := dto.ChatCompletionsStreamResponse{
  166. Object: "chat.completion.chunk",
  167. Created: common.GetTimestamp(),
  168. Model: "dify",
  169. }
  170. var choice dto.ChatCompletionsStreamResponseChoice
  171. if strings.HasPrefix(difyResponse.Event, "workflow_") {
  172. if constant.DifyDebug {
  173. text := "Workflow: " + difyResponse.Data.WorkflowId
  174. if difyResponse.Event == "workflow_finished" {
  175. text += " " + difyResponse.Data.Status
  176. }
  177. choice.Delta.SetReasoningContent(text + "\n")
  178. }
  179. } else if strings.HasPrefix(difyResponse.Event, "node_") {
  180. if constant.DifyDebug {
  181. text := "Node: " + difyResponse.Data.NodeType
  182. if difyResponse.Event == "node_finished" {
  183. text += " " + difyResponse.Data.Status
  184. }
  185. choice.Delta.SetReasoningContent(text + "\n")
  186. }
  187. } else if difyResponse.Event == "message" || difyResponse.Event == "agent_message" {
  188. if difyResponse.Answer == "<details style=\"color:gray;background-color: #f8f8f8;padding: 8px;border-radius: 4px;\" open> <summary> Thinking... </summary>\n" {
  189. difyResponse.Answer = "<think>"
  190. } else if difyResponse.Answer == "</details>" {
  191. difyResponse.Answer = "</think>"
  192. }
  193. choice.Delta.SetContentString(difyResponse.Answer)
  194. }
  195. response.Choices = append(response.Choices, choice)
  196. return &response
  197. }
  198. func difyStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {
  199. var responseText string
  200. usage := &dto.Usage{}
  201. var nodeToken int
  202. helper.SetEventStreamHeaders(c)
  203. helper.StreamScannerHandler(c, resp, info, func(data string) bool {
  204. var difyResponse DifyChunkChatCompletionResponse
  205. err := json.Unmarshal([]byte(data), &difyResponse)
  206. if err != nil {
  207. common.SysLog("error unmarshalling stream response: " + err.Error())
  208. return true
  209. }
  210. var openaiResponse dto.ChatCompletionsStreamResponse
  211. if difyResponse.Event == "message_end" {
  212. usage = &difyResponse.MetaData.Usage
  213. return false
  214. } else if difyResponse.Event == "error" {
  215. return false
  216. } else {
  217. openaiResponse = *streamResponseDify2OpenAI(difyResponse)
  218. if len(openaiResponse.Choices) != 0 {
  219. responseText += openaiResponse.Choices[0].Delta.GetContentString()
  220. if openaiResponse.Choices[0].Delta.ReasoningContent != nil {
  221. nodeToken += 1
  222. }
  223. }
  224. }
  225. err = helper.ObjectData(c, openaiResponse)
  226. if err != nil {
  227. common.SysLog(err.Error())
  228. }
  229. return true
  230. })
  231. helper.Done(c)
  232. if usage.TotalTokens == 0 {
  233. usage = service.ResponseText2Usage(c, responseText, info.UpstreamModelName, info.GetEstimatePromptTokens())
  234. }
  235. usage.CompletionTokens += nodeToken
  236. return usage, nil
  237. }
  238. func difyHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {
  239. var difyResponse DifyChatCompletionResponse
  240. responseBody, err := io.ReadAll(resp.Body)
  241. if err != nil {
  242. return nil, types.NewError(err, types.ErrorCodeBadResponseBody)
  243. }
  244. service.CloseResponseBodyGracefully(resp)
  245. err = json.Unmarshal(responseBody, &difyResponse)
  246. if err != nil {
  247. return nil, types.NewError(err, types.ErrorCodeBadResponseBody)
  248. }
  249. fullTextResponse := dto.OpenAITextResponse{
  250. Id: difyResponse.ConversationId,
  251. Object: "chat.completion",
  252. Created: common.GetTimestamp(),
  253. Usage: difyResponse.MetaData.Usage,
  254. }
  255. choice := dto.OpenAITextResponseChoice{
  256. Index: 0,
  257. Message: dto.Message{
  258. Role: "assistant",
  259. Content: difyResponse.Answer,
  260. },
  261. FinishReason: "stop",
  262. }
  263. fullTextResponse.Choices = append(fullTextResponse.Choices, choice)
  264. jsonResponse, err := json.Marshal(fullTextResponse)
  265. if err != nil {
  266. return nil, types.NewError(err, types.ErrorCodeBadResponseBody)
  267. }
  268. c.Writer.Header().Set("Content-Type", "application/json")
  269. c.Writer.WriteHeader(resp.StatusCode)
  270. c.Writer.Write(jsonResponse)
  271. return &difyResponse.MetaData.Usage, nil
  272. }