AI chatbot call center: Telegram call in (production-ready, part 1a)
UnknownUnknown. This template uses community nodes we can't verify statically. Uses community nodes that can't be verified statically (unverified community packages are off by default in 3.0). Static check, last run 2026-10-19.
Community nodes
These come from third-party npm packages. In n8n 3.0 unverified community packages are off by default (N8N_UNVERIFIED_PACKAGES_ENABLED=false).
n8n-nodes-google-speech(Google STT)
Nodes used (12)
- Telegram Trigger
n8n-nodes-base.telegramTrigger - Chat Trigger
@n8n/n8n-nodes-langchain.chatTrigger - Edit Fields (Set)
n8n-nodes-base.set - Switch
n8n-nodes-base.switch - Telegram
n8n-nodes-base.telegram - Extract from File
n8n-nodes-base.extractFromFile - Google Speech
n8n-nodes-google-speech.googleSpeech - If
n8n-nodes-base.if - Code
n8n-nodes-base.code - Redis
n8n-nodes-base.redis - Postgres
n8n-nodes-base.postgres - Execute Sub-workflow
n8n-nodes-base.executeWorkflow
What it does
Workflow Name: 🤙 Telegram Call In Template was created in n8n v1.90.2 Skill Level: High Categories: n8n, Chatbot Stacks - Chat Trigger node - Telegram Trigger node - Redis node - Postgres node - Execute Sub-workflow - If node, Code node, Edit Fields (Set), Extract From File Prerequisite - Community nodes: n8n-nodes-google-speech - Sub-workflow: Demo Call Back - Sub-workflow: Demo Call Center Production Features - Scaling Design for n8n Queue mode in production environment - Optional Member Data from external Database with Caching Mechanism. - Optional Voice Message to Text Message via Google STT API with Multi-Languages support. - Testing Flow with or without dependance on other workflow. - Error Management What this workflow does? This is a n8n Telegram Call In Workflow. It will wait for message from Telegram bot and sent to the Call Center to process. How it works - The Telegram Trigger node will wait for the message from the Telegram bot. - When message is received, it will first check for the matching Member from the PostgreSQL database. - Then determine if it is a Text message or Voice message. - For voice message, use the Google Speech to Text API to transcript it into text. - Finally, pass the text to the next flow, i.e. the Call Center. Set up instructions n8n-nodes-google-speech 1. Pull and Set up the required SQL from our Github repository. 2. Go User > Settings > Community nodes, install n8n-nodes-google-speech node 1. Follow https://www.npmjs.com/package/n8n-nodes-google-speech 2. Setup the Google STT node 3. Create you Telegram credentials, refer to n8n integration documentation for more information. 4. Select your Credentials in Telegram Trigger 5. Create you Redis credentials, refer to n8n integration documentation for more information. 6. Select your Credentials in Member Cache and Save Member Cache. 7. Create you Postgres credentials, refer to n8n integration documentation for more information. 8. Select your Credentials in Load Member Data. 9. FOR TEST ONLY. Enable the Telegram Test Output for testing from the Test Input. 10. Remember to activate this workflow for incoming message. How to adjust it to your needs 1. By default, this template will use the sysmember table for member information, you could change it for your own design. 2. The demo implementation does not include failed member loading situation, you should implement based on your needs, e.g. if isactive is not true, do… 3. You can replace the sub-workflow Demo Call Back and Demo Call Center to another flow as needs.