Whataspp integration

How To
1

Hello! I’ve set up an integration in make.com with timelines.ai where the first module in timelines.ai monitors new incoming WhatsApp messages. Then, it sends the incoming message to a chatGPT agent, which generates a reply and sends it to another timelines.ai module for delivery.

There is one issue – the GPT agent doesn’t retain even the previous message from the client. With every new message from the same client, the agent responds as if it’s the first interaction. Why does this happen, and how can it be fixed?

The second point is that I need the agent to respond only to unread messages in WhatsApp. If the message is already read (for example, if a human is currently working with the client), the bot should remain silent and not interfere in the conversation. In timelines.ai, there is a module that checks for new outgoing messages in the chat, but it has no input point. This means I can’t set it up as a conditional check to stop the bot from responding if there are sent messages. Alternatively, is there a way to configure timelines.ai to activate only for unread messages? How can this be done?

image
image1286×809 69.9 KB

2

For the issue with the loss of context with the GPT bot, make sure you are using a thread-id that doesn’t change per message. It has to be something that remains unique for the duration of the conversation and that doesn’t repeat within other conversations. The contact’s phone number is usually enough in these cases.

thread-id
thread-id1258×1116 139 KB

The second point will need a kill switch in your Make scenario. For example, you could check if any reply came from the WhatsApp application and disable the bot in those cases.

I don’t know exactly if Timelines lets you do this, but with 2Chat it’s possible.