The Mezmo AI Assistant is designed to help you make faster, smarter decisions with your telemetry data. It works best when you guide it with clear, scoped prompts.
✅ Optimized For
Troubleshooting production issues
Detecting patterns across telemetry data
Explaining anomalies in plain language
Summarizing large volumes of logs
Identifying likely root causes faster
🧠 How to Use the AI Assistant Effectively
Use the built-in prompt suggestions to get started quickly.
Start broad, then narrow your prompt as you refine the investigation.
Think of the AI as a teammate — ask questions as you would with an engineer.
Include time, scope, and intent in your prompt for best results.
🛠️ It can also generate Mermaid visualizations
For example, you can ask:
“Show me a pie chart of service usage for app:restaurant-app in the last 24h”
pie showData
title _service for app:restaurant-app (last 24h)
"restaurant-performance" : 18269007
"metrics" : 4035581 "events" : 2200573
"performance" : 2200502
"access" : 373064
"app" : 27167
"server" : 24
"errors" : 5
🎯 Prompt Structure: What to Include (in order)
Element | Description |
1. Goal | What you're trying to achieve: RCA, reduce noise, fix errors, etc. |
2. Scope | Pipeline ID, app/service name, environment, cluster, etc. |
3. Time Window | Use UTC, e.g. “from 12:00 to 12:30” or “last 15 minutes” |
4. Symptom/Impact | E.g., “timeouts”, “dropped events”, “5xx errors” |
5. Known Filters | E.g., level:error, trace_id, host, customer ID (avoid PII) |
6. Desired Output | E.g., top patterns, timeline, likely root cause, next steps |
🧾 Sample Prompts
🔍 Root Cause Analysis (RCA)
“Run RCA for auth-service from 12:00 UTC to 12:30 UTC.
Symptoms: timeouts and 5xx errors.
Return: trigger → impact → top log clusters → recommended action.”
🔄 What Changed?
“Compare checkout-service logs between
09:00–10:00 UTC (baseline) and 12:00–13:00 UTC (incident).
What new errors or increases occurred?”
📉 Reduce Log Noise / Cost
“Identify top noisy logs in prod during last 1h.
Group by app and message. Suggest filters or redactions (include risks).”
🔁 Detect Duplicate Logs
“Check for duplicate logs in orders-service (prod) during 10:00–10:30 UTC.
Show dominant patterns and causes like retries or multi-shipping.”
📐 Build a Query
“Help write a Mezmo query to find timeouts in cart-service (prod)
excluding known retry noise. Fields: status_code, latency_ms, trace_id.”
🔎 Explore What’s Happening
“Show me the top services by volume in prod during last 30 minutes.”
Current documentation and safe use
For feature availability, setup, and product-specific limits, use the current Mezmo documentation. The examples above are investigation patterns, not guarantees about accuracy or supported capabilities.
Start with a focused time window and a clearly identified service, account, or pipeline. When results are empty or unexpected, verify your filters, timestamps, permissions, and incoming data before drawing conclusions.
Review generated queries and proposed remediation before applying changes. Do not include API tokens, passwords, or unnecessary personal information in prompts, screenshots, or support requests.
For help, open Support in the Mezmo app. You can review existing conversations and tickets or deliberately start a new question.