llama-farm /llamafarm:ml-status - ML Services Health Check
Check the health and status of all LlamaFarm ML components — runtime, anomaly models, classifiers, streaming detectors, and data buffers.
Usage
/llamafarm:ml-status # Full ML status report
/llamafarm:ml-status --component anomaly # Check anomaly models only
/llamafarm:ml-status --component classify # Check classifiers only
/llamafarm:ml-status --component buffers # Check Polars buffers only
What This Command Does
- Checks runtime health — verifies the Universal Runtime is responding
- Lists anomaly models — trained batch models and their metadata
- Lists streaming detectors — active streaming anomaly detectors
- Lists classifiers — trained text classification models
- Lists data buffers — active Polars buffers and their sizes
- Presents formatted report — consolidated view of all ML resources
Implementation
When the user runs /llamafarm:ml-status, follow these steps:
Step 1: Check Universal Runtime Health
curl -s http://localhost:14345/health | jq .
Expected healthy response:
{
"status": "healthy",
"components": {
"server": "healthy",
"celery": "healthy",
"database": "healthy"
},
"version": "0.x.x"
}
If the health check fails, report the runtime as unavailable and suggest running /llamafarm:start --runtime.
Step 2: List Anomaly Models
curl -s http://localhost:14345/v1/ml/anomaly/models | jq .
Step 3: List Streaming Detectors
curl -s http://localhost:14345/v1/ml/anomaly/detectors | jq .
Step 4: List Classifiers
curl -s http://localhost:14345/v1/classifier/models | jq .
Step 5: List Polars Buffers
curl -s http://localhost:14345/v1/ml/buffers | jq .
Step 6: Present Comprehensive Status
LlamaFarm ML Status Report
===========================
Runtime: HEALTHY (http://localhost:14345)
Anomaly Models (batch):
Name Backend Version Trained
sensor_model iforest 3 2024-01-14 10:30
fraud_detector autoencoder 1 2024-01-13 15:22
Streaming Detectors:
Name Backend Window Samples Seen
live_sensor ecod 1000 4,521
network_monitor hbos 500 12,034
Classifiers:
Name Base Model Labels Accuracy
support_tickets distilbert-base-uncased 5 92.3%
doc_classifier roberta-base 3 96.1%
Polars Buffers:
Name Rows Columns Memory
sensor_data 10,000 12 1.2 MB
transactions 50,000 8 3.8 MB
Summary:
Anomaly Models: 2
Streaming Detectors: 2
Classifiers: 2
Polars Buffers: 2
Overall: All systems operational
Component Details
Runtime Health
| Status | Meaning |
|---|---|
healthy |
Runtime responding, all ML endpoints available |
degraded |
Runtime running but some backends unavailable |
failed |
Runtime not responding |
Model States
| State | Meaning |
|---|---|
ready |
Model loaded and ready for inference |
training |
Model currently being trained |
error |
Model failed to load or train |
Troubleshooting Output
When issues are detected:
LlamaFarm ML Status Report
===========================
Runtime: DEGRADED
Issues Detected:
1. Universal Runtime Not Responding
- Health endpoint returned connection refused
- Fix: lf runtime start or /llamafarm:start --runtime
2. Anomaly Model Load Error
- Model "sensor_model" failed to load: missing backend dependency
- Fix: Install required backend: pip install pyod
Recommendations:
- Run /llamafarm:logs --service runtime to see error details
- Run /llamafarm:start --runtime to restart the ML runtime
Related Commands
/llamafarm:status- Check overall service health/llamafarm:anomaly- Anomaly detection workflows/llamafarm:classify- Text classification workflows/llamafarm:ocr- OCR text extraction