N8N Software Automation Help

Hi!

I’m running into a very strange issue on ---------- with a Telegram ---------- workflow and would like to see if you could help fix it. Please let me know if you are interested.

I'm using a template from n8n: ----------

My Goal:
The workflow is a RAG (Retrieval-Augmented Generation) chatbot. It uses a Telegram Trigger, an AI Agent node with Google Gemini and a Pinecone Vector Store tool, and a final Telegram (sendMessage) node to reply to the user.

The Problem
My workflow now executes perfectly from start to finish. When I test it live from my Telegram app, the n8n execution log shows a “Succeeded” status with a long execution time (approx. 30 seconds). The execution path shows a **green checkmark on every single node, including the final Telegram (sendMessage) node.

This proves that n8n is successfully sending the request to the Telegram API and receiving a “200 OK” success response. However, the reply message never arrives in my Telegram client.

My Setup:
n8n Version: 1.95.3 (Self Hosted on Hostinger)
Final Workflow Pattern:
1. Telegram Trigger connects to the AI Agent.
2. Telegram Trigger also connects to Input 1 of a Merge node.
3. AI Agent connects to Input 2 of the Merge node.
4. The Merge node connects to the final Telegram (sendMessage) node.

What I Have Already Ruled Out:
Blocked Bot:I have triple-checked that my bot is not blocked in my Telegram app.
Mismatched Credentials: I have verified that the exact same Telegram credential is used in both the Telegram Trigger and the final Telegram (sendMessage) node.
Parse Mode: I have set the Parse Mode in the final Telegram node to None. The issue persists.
Workflow Logic: We have isolated and tested every component. The AI Agent correctly generates a text output. The workflow logic appears to be perfect.

Execution Screenshot: ----------

Workflow Screenshot: ----------

Please let me know if this is something that you can help with, the cost and when you can complete it.

There will be some additional work that I would like your help with on this automation such as isolating the AI to only give outputs on the documents...etc. First let's tackle one thing at a time. Thank you!