HEIMR.AI
AI-Powered Load Test Analysis
Explainable Root Cause Analysis for Your Performance Tests
AI-Powered Load Test Analysis
Explainable Root Cause Analysis for Your Performance Tests
December 7, 2025 • 1:00 PM CET
Analyzes JMeter, k6, Gatling, and Locust test results automatically
Combines 6+ signals: latency anomalies, errors, CPU, memory, logs, and traces
Local LLM (Llama 3.1 / Ministral) generates root cause analysis instantly
No black-box ML - every anomaly has clear statistical reasoning
Runs completely local with Ollama - your data never leaves your machine
O(n) complexity - no training required, works out of the box
Heimr.ai correlates your load test results with observability data (Prometheus, Loki, Tempo) to identify performance bottlenecks. Using statistical anomaly detection and AI-powered analysis, it generates comprehensive markdown reports with actionable recommendations.
✨ Think of it as having a senior performance engineer analyze your tests 24/7
╔═══════════════════╗
║ Load Test Files ║
║ JMeter│k6│Gatling ║
║ Locust ║
╚═════════┬═════════╝
│
▼
┌─────────┐
│ PARSERS │
└────┬────┘
│
▼
┌──────────────┐ ┌─────────────────────┐
│ ANOMALY │────▶│ OBSERVABILITY DATA │
│ DETECTOR │ │ Prometheus│Loki │
│ 6 Signals │ │ Tempo │
└──────┬───────┘ └─────────────────────┘
│ │
└───────────┬───────────┘
▼
┌────────────────┐
│ LLM CLIENT │
│ Llama / Claude │
└────────┬───────┘
▼
┌────────────────┐
│ MARKDOWN REPORT│
│Root Cause + Fix│
└────────────────┘
$ heimr analyze results.jtl --explain
Analyzing results.jtl (jmeter)...
✅ Found 100 requests
⚠️ Detected 7 anomalies
📊 Fetching Prometheus metrics...
📝 Fetching Loki logs...
🔍 Fetching Tempo traces...
❌ FAILED
Reasons: Error Rate: 39.00%, Anomalies: 7, Error/Warn Logs: 8
🤖 AI Analysis:
Root Cause: Database saturation caused by
unoptimized queries. Recommend adding indexes
and implementing connection pooling.
✅ Report saved to: analysis_report.md
█
Senior Performance Engineer & ML Enthusiast
Performance engineer passionate about making complex systems faster and more reliable. Built Heimr.ai after spending too many late nights manually analyzing load test results and correlating metrics across multiple observability systems. Believes in explainable AI over black-box solutions and local-first tools that respect your privacy.
When not optimizing performance or training models, you'll find me exploring chaos engineering, contributing to open source, or experimenting with the latest LLMs.
Connect with me professionally
@juan-estevez-castilloDrop me a message
jd.estevezcastillo@gmail.comInterested in beta testing? Want to collaborate? Let's talk!