See llms.txt for all machine-readable content.

Back to Templates

Score chatbot replies, log reviews, and alert your team with Mistral and Telegram

Created by

Created by: DARGE JET || dargejet
DARGE JET

Last update

Last update 21 hours ago

Categories

Quick overview

This workflow reviews each chatbot answer using Mistral Medium, stores an anonymized quality score in an n8n Data Table, and sends a Telegram alert when the answer is poor or the client seems unhappy or requests a human.

How it works

  1. Starts either when you manually execute the workflow (using a sample conversation) or when it is called by your chatbot as a sub-workflow.
  2. Merges your incoming chat data with configuration settings like bot rules, Telegram chat ID, alert threshold, and audit rate.
  3. Sends the conversation context, client message, bot answer, and your rules to Mistral Medium to score accuracy, rule-following, tone, and handover, and to classify the client’s mood and whether they want a human.
  4. Parses the model’s JSON response, calculates the overall score, decides whether to keep a short reason/excerpt for auditing, and flags the interaction for alerting (including handling model/parse errors).
  5. Writes one review row (without chat identifiers) to the n8n Data Table.
  6. If the item is flagged, sends a formatted alert to your team in Telegram with the score, mood, handover status, and the stored reason/excerpt.

Setup

  1. Add a Mistral Cloud credential for the Mistral Medium chat model.
  2. Create an n8n Data Table named ai_reviews and ensure it includes fields for score, mood, status, reason, excerpt, lang, wants_human, handed_over, and alert.
  3. Add a Telegram bot credential, add the bot to your team group, and set the correct team_chat_id in the Configuration step.
  4. Update the rules, alert_max_score, and audit_rate values in the Configuration step to match your chatbot policy and alerting preferences.
  5. If calling from your chatbot, configure it to pass question, answer, summary, lang, and handed_over into this workflow’s Execute Workflow trigger.

Requirements

  • Mistral API key (Mistral Medium)
  • Telegram bot that is a member of your team group
  • n8n with Data tables (self-hosted or Cloud)

Customization

  • Change alert_max_score and audit_rate in the Configuration node
  • Replace Mistral Medium with any other chat model node
  • Add a scheduled cleanup that deletes old rows according to your privacy policy

Additional info

Built for our own Telegram AI assistant and used in production. Tested on self-hosted n8n 2.x.