How to distinguish "fromMe" messages sent via API vs. WhatsApp Web in Evolution API v2.x?

Questions
, ,
1

Hi everyone,

I’m building an automation in n8n using Evolution API v2.3.7 to handle lead qualification for a law firm. I have a logic that silences the AI Agent for 30 minutes if the lawyer intervenes manually (sending a message from her phone or WhatsApp Web).

The Setup:

  • Trigger: Webhook listening to messages.upsert.

  • Database: Google Sheets/Postgres tracks the last_human_interaction timestamp.

  • Logic: If fromMe is true, I update the timestamp. If a new message arrives and the last interaction was < 30 mins ago, the bot remains silent.

The Issue: The Evolution API sends a webhook for every message, including the ones sent by the n8n bot itself via the API.

  • Bot messages (via API) arrive as: fromMe: true and source: web.

  • Manual messages (Lawyer via WhatsApp Web) arrive as: fromMe: true and source: web.

Because both have the exact same source, the bot records its own responses as “Human Interactions,” triggering its own silence logic and effectively “suiciding” after the first reply.

The Question: Is there a reliable way in the messages.upsert payload to distinguish between a message sent specifically via the API and a message sent manually from the WhatsApp Web interface when the source field is identical?

I’ve tried filtering by source, but since the lawyer uses WhatsApp Web, it’s not a unique identifier.

Environment:

  • Evolution API Version: 2.3.7

  • n8n Version: 2.14.2

  • Instance Type: Typebot/n8n integration.

    image
    image828×354 30.6 KB

I have disabled the nodes related to the solution because they were creating problems.

Has anyone found a specific metadata field or a header that identifies API-originated messages?

Thanks in advance!

2

Hi Since you can’t rely on the source field, the best workaround is to tag your bot’s outgoing messages with a custom identifier before sending them. You can add a unique prefix or invisible character to the message body, or store the message ID of every message your bot sends in your database. Then when the webhook fires with fromMe: true, check if that message ID already exists in your records… if it does, it’s a bot message and you skip the timestamp update. If it doesn’t exist, it’s the lawyer and you trigger the silence logic.

2 Likes
3

thank you, this really worked, i thought about that possibily, but your comment made me implement something similar, following that logic. thank you, have a nice day.

1 Like
4

Glad it worked out, good luck with the rest of the build :+1:

1 Like
5

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.