The attached image shows i ma able to have a working chatbot embedded in external site. It can send email, pull user’s calendar.
But my biggest struggle is, how could the bot being embedded in 3rd party website, still able to tool returned based on the logged user? Currently i noticed all tutorial i visited, are putting a static user. In actual cuase case, the bot is expose as a chat app that “know the user/employee” and reply based on that particular user (email) and not supposed to be a generic email.
Hope can get some inspiration… thanks
spec
- Self hosted / macOS / Docker
- version: 1.81.4
- DB: default
even we could “piggy bag” in our encrypted email (Chat Trigger node common issues | n8n Docs), it’s still a challenge if we wanna deliver for public. (refer screenshot),
- Gmail OAuth 2.0 is authenticated to my own GCP. Meaning if i invite other Google user, it’s unable to summarize their email/calendar due to security reason
- The User (email) is channeled from my .html (simulate 3rd party app), but the $json.metadata.User is not resolve, meaning i can’t dynamically set the email “To”, but can only hardcode instead.
Hope to gather some feedback for whoever come across. Thanks in million
yeah it does not seem like one can use N8N for user specefic AI interactions unless you create a metadata for a chat memory function and create a flow to store the chat under the user’s metadata. the problem is that each message alone will count as one execution and not each chat session. THe only way to have it is to run it locally if you want to not sell a kidney for user facing products via N8N
1 Like
