🚸 (openai) Parse stream on client to correctly handle errors

packages/embeds/js/src/utils/executeClientSideActions.ts

@@ -7,14 +7,16 @@ import { executeSetVariable } from '@/features/blocks/logic/setVariable/executeS
 import { executeWait } from '@/features/blocks/logic/wait/utils/executeWait'
 import { executeWebhook } from '@/features/blocks/integrations/webhook/executeWebhook'
 import { ClientSideActionContext } from '@/types'
-import type { ChatReply } from '@typebot.io/schemas'
+import type { ChatReply, ReplyLog } from '@typebot.io/schemas'
 
 export const executeClientSideAction = async (
   clientSideAction: NonNullable<ChatReply['clientSideActions']>[0],
   context: ClientSideActionContext,
   onStreamedMessage?: (message: string) => void
 ): Promise<
-  { blockedPopupUrl: string } | { replyToSend: string | undefined } | void
+  | { blockedPopupUrl: string }
+  | { replyToSend: string | undefined; logs?: ReplyLog[] }
+  | void
 > => {
   if ('chatwoot' in clientSideAction) {
     return executeChatwoot(clientSideAction.chatwoot)
@@ -35,11 +37,22 @@ export const executeClientSideAction = async (
     return executeSetVariable(clientSideAction.setVariable.scriptToExecute)
   }
   if ('streamOpenAiChatCompletion' in clientSideAction) {
-    const text = await streamChat(context)(
+    const { error, message } = await streamChat(context)(
       clientSideAction.streamOpenAiChatCompletion.messages,
       { onStreamedMessage }
     )
-    return { replyToSend: text }
+    if (error)
+      return {
+        replyToSend: undefined,
+        logs: [
+          {
+            status: 'error',
+            description: 'Failed to stream OpenAI completion',
+            details: JSON.stringify(error, null, 2),
+          },
+        ],
+      }
+    return { replyToSend: message }
   }
   if ('webhookToExecute' in clientSideAction) {
     const response = await executeWebhook(clientSideAction.webhookToExecute)