fix: #3673 - API responds with Request body is too large

core/src/node/api/restful/helper/builder.test.ts

@@ -236,6 +236,47 @@ describe('builder helper functions', () => {
       })
     })
 
+    it('should return the error on status not ok', async () => {
+      const request = { body: { model: 'model1' } }
+      const mockSend = jest.fn()
+      const reply = {
+        code: jest.fn().mockReturnThis(),
+        send: jest.fn(),
+        headers: jest.fn().mockReturnValue({
+          send: mockSend,
+        }),
+        raw: {
+          writeHead: jest.fn(),
+          pipe: jest.fn(),
+        },
+      }
+
+      ;(existsSync as jest.Mock).mockReturnValue(true)
+      ;(readdirSync as jest.Mock).mockReturnValue(['file1'])
+      ;(readFileSync as jest.Mock).mockReturnValue(
+        JSON.stringify({ id: 'model1', engine: 'openai' })
+      )
+
+      // Mock fetch
+      const fetch = require('node-fetch')
+      fetch.mockResolvedValue({
+        status: 400,
+        headers: new Map([
+          ['content-type', 'application/json'],
+          ['x-request-id', '123456'],
+        ]),
+        body: { pipe: jest.fn() },
+        text: jest.fn().mockResolvedValue({ error: 'Mock error response' }),
+      })
+      await chatCompletions(request, reply)
+      expect(reply.code).toHaveBeenCalledWith(400)
+      expect(mockSend).toHaveBeenCalledWith(
+        expect.objectContaining({
+          error: 'Mock error response',
+        })
+      )
+    })
+
     it('should return the chat completions', async () => {
       const request = { body: { model: 'model1' } }
       const reply = {

core/src/node/api/restful/helper/builder.ts

@@ -42,15 +42,15 @@
         try {
           return JSON.parse(result)
         } catch (err) {
           console.error(err)
         }
       })
       .filter((e: any) => !!e)
 
     return modelData
   } catch (err) {
     console.error(err)
     return []
   }
 }
 
@@ -58,7 +58,7 @@
   if (existsSync(path)) {
     return readFileSync(path, 'utf-8')
   } else {
     return undefined
   }
 }
 
@@ -97,7 +97,7 @@
       deleted: true,
     }
   } catch (ex) {
     console.error(ex)
   }
 }
 
@@ -113,8 +113,8 @@
 
     const messageFilePath = join(threadDirPath, messageFile)
     if (!existsSync(messageFilePath)) {
       console.debug('message file not found')
       return []
     }
 
     const lines = readFileSync(messageFilePath, 'utf-8')
@@ -128,8 +128,8 @@
     })
     return messages
   } catch (err) {
     console.error(err)
     return []
   }
 }
 
@@ -172,7 +172,7 @@
     await writeFileSync(threadJsonPath, JSON.stringify(updatedThread, null, 2))
     return updatedThread
   } catch (err) {
     return {
       error: err,
     }
   }
@@ -202,7 +202,7 @@
     await writeFileSync(threadJsonPath, JSON.stringify(updatedThread, null, 2))
     return updatedThread
   } catch (err) {
     return {
       message: err,
     }
   }
@@ -247,7 +247,7 @@
     appendFileSync(threadMessagePath, JSON.stringify(threadMessage) + '\n')
     return threadMessage
   } catch (err) {
     return {
       message: err,
     }
   }
@@ -268,7 +268,7 @@
 
   const directoryPath = join(getJanDataFolderPath(), 'models', modelId)
   if (!existsSync(directoryPath)) {
     mkdirSync(directoryPath)
   }
 
   // path to model binary
@@ -353,8 +353,10 @@
     body: JSON.stringify(request.body),
   })
   if (response.status !== 200) {
-    console.error(response)
-    reply.code(400).send(response)
+    // Forward the error response to client via reply
+    const responseBody = await response.text()
+    const responseHeaders = Object.fromEntries(response.headers)
+    reply.code(response.status).headers(responseHeaders).send(responseBody)
   } else {
     reply.raw.writeHead(200, {
       'Content-Type': request.body.stream === true ? 'text/event-stream' : 'application/json',

server/index.ts

@@ -67,6 +67,11 @@ export const startServer = async (configs?: ServerConfig): Promise<boolean> => {
     // Initialize Fastify server with logging
     server = fastify({
       logger: new Logger(),
+      // Set body limit to 100MB - Default is 1MB
+      // According to OpenAI - a batch input file can be up to 100 MB in size
+      // Whisper endpoints accept up to 25MB
+      // Vision endpoints accept up to 4MB
+      bodyLimit: 104_857_600
     })
 
     // Register CORS if enabled