Diagnostics

This page covers advanced diagnostic procedures for investigating PRX issues that are not resolved by the basic troubleshooting steps.

Diagnostic Commands

prx doctor

The comprehensive health check:

prx doctor

Output includes:

• Configuration validation results
• Provider connectivity tests
• System dependency checks
• Resource usage summary

prx debug

Enable debug-level logging for detailed operation traces:

PRX_LOG=debug prx daemon

Or set in config:

[observability]
log_level = "debug"

prx info

Display system information:

prx info

Shows:

• PRX version and build info
• OS and architecture
• Configured providers and their status
• Memory backend type and size
• Plugin count and status

Log Analysis

PRX logs are structured JSON (when log_format = "json"). Key fields to look for:

Field	Description
level	Log level (debug, info, warn, error)
target	Rust module path
session_id	Associated session ID
provider	LLM provider involved
duration_ms	Operation duration
error	Error details (if applicable)

Network Diagnostics

Test provider connectivity:

# Test Anthropic API
prx provider test anthropic

# Test all configured providers
prx provider test --all

# Check network from sandbox
prx sandbox test-network

Performance Profiling

Enable the metrics endpoint and use Prometheus/Grafana for performance analysis:

[observability.metrics]
enabled = true
bind = "127.0.0.1:9090"

Key metrics to monitor:

• prx_llm_request_duration_seconds -- LLM latency
• prx_sessions_active -- concurrent sessions
• prx_memory_usage_bytes -- memory consumption

Related Pages

• Troubleshooting Overview
• Observability
• Prometheus Metrics