Observability: Health, Monitoring & Logs¶

This guide covers the tools and procedures for monitoring the Cloudflare RAG service.

1. Health Checks¶

Health checks are the first line of defense for detecting issues.

API-Side Health Probes¶

The Labeeb API provides endpoints to check the health of its Cloudflare dependencies from its own perspective.

Worker-Side Health¶

• Endpoint: /rag/health
• Deep Check: Use ?deep=1 to include a quick embedding timing test, which verifies the Workers AI binding is functional.

2. Logs¶

Real-time logs are essential for debugging.

• Command: Use wrangler tail to stream live logs from your worker.

  # Tail logs for a specific environment
  wrangler tail --env stage
  

• AI Gateway Logs: View detailed logs and analytics in the Cloudflare dashboard under AI → AI Gateway → Logs.

3. Monitoring & Dashboards¶

AI Gateway Dashboard¶

The AI Gateway dashboard is the primary tool for monitoring cost, latency, and errors related to third-party LLM providers.

• Location: AI → AI Gateway → Analytics
• Key Metrics to Watch:
  • Cache Hit Ratio: Enable Cache responses in the gateway settings to reduce costs and latency. A high cache hit ratio is desirable.
  • Errors: Monitor for spikes in error rates, which could indicate issues with an upstream provider.
  • Latency: Track p95 latency to ensure providers are meeting performance expectations.

Vectorize¶

Use the Wrangler CLI to inspect the status and metadata of your Vectorize indexes.

# List all Vectorize indexes in your account
wrangler vectorize list

# List the metadata configuration for a specific index
wrangler vectorize list-metadata-index labeeb-articles-dev

Server-Timing Header¶

The /rag/query endpoint includes a Server-Timing response header, which provides a detailed breakdown of the time spent in each stage of the RAG pipeline (e.g., embed, search, rerank). This is invaluable for performance analysis.