Michael Brenndoerfer | Data & AI, Private Equity, Technology

Michael Brenndoerfer

Monitoring and Reliability › Deeper Health Checks

Michael BrenndoerferAugust 20, 20252 min read

AI Agent HandbookAll chapters

Part 14 of 15Deployment and Operations - From Prototype to Production

Chapter 40Monitoring and Reliability › Deeper Health Checks

Reading section 4 of 14 · Deeper Health Checks · 2 minSection 4 of 14 · 2 min

Section 4 of 14 · 2 min read

Monitoring and Reliability › Deeper Health ChecksOne section · Sections in this chapter4 / 14 ↓

Sections in this chapter14

A basic health check tells you the web server is responding, but your agent depends on more than just the web framework. It needs to call the language model API, access any databases or memory stores, and use external tools. A more thorough health check verifies these dependencies too.

Here's an enhanced health check that tests key components:

In[2]:

Code

@app.route('/health', methods=['GET'])
def health_check():
    """Comprehensive health check"""
    health_status = {
        "status": "healthy",
        "timestamp": time.time(),
        "checks": {}
    }
    
    # Check if we can reach the LLM API
    try:
        client = Anthropic(api_key=os.getenv("ANTHROPIC_API_KEY"))
        # Make a minimal API call to verify connectivity
        response = client.messages.create(
            model="claude-sonnet-4-5",
            max_tokens=10,
            messages=[{"role": "user", "content": "test"}]
        )
        health_status["checks"]["llm_api"] = "ok"
    except Exception as e:
        health_status["checks"]["llm_api"] = f"error: {str(e)}"
        health_status["status"] = "degraded"
    
    # Check database connectivity
    try:
        # Test database connection
        db.execute("SELECT 1")
        health_status["checks"]["database"] = "ok"
    except Exception as e:
        health_status["checks"]["database"] = f"error: {str(e)}"
        health_status["status"] = "degraded"
    
    return jsonify(health_status)

This health check tests the critical dependencies. If the LLM API is unreachable or the database is down, the health check reports "degraded" status. You know the agent is running, but it can't function properly.

Be careful with comprehensive health checks, though. You don't want the health check itself to be expensive or slow. Keep the checks lightweight. A minimal API call to verify connectivity is fine, but don't run a full agent query as part of the health check.