API Health Checker

Comprehensive API health assessment engine with scoring, multi-dimensional evaluation, and actionable insights. Perfect for uptime monitoring, SLA verification, and automated health checks.

Now with GenAI/LLM API monitoring – supports OpenAI, Anthropic, Google AI, and Azure OpenAI.

Features

• Health Scoring (0-100) – Quantitative assessment of each endpoint's health
• Multi-Dimensional Evaluation – Availability, performance, validation, reliability
• Actionable Insights – Deterministic recommendations for operators
• Performance Rating – fast / acceptable / slow classification
• Configurable Scoring – Custom thresholds and weights
• Retry Logic – Exponential backoff for transient failures
• Fault Tolerant – Individual failures don't crash the run

GenAI/LLM Features

• Multi-Provider Support – OpenAI, Anthropic, Google AI, Azure OpenAI, Custom
• Token Usage Tracking – Input/output/total tokens per request
• Rate Limit Detection – Automatic detection and retry on 429 errors
• Content Filtering – Detects safety filter activations
• Time-to-First-Token – Streaming latency measurement (coming soon)

Health Scoring System

Each endpoint receives a health score from 0-100 based on four dimensions:

Dimension	Weight	Description
Availability	40%	Endpoint reachable + correct status code
Performance	30%	Latency rating (fast=100%, acceptable=67%, slow=33%)
Validation	20%	Response validation passed
Reliability	10%	Retry behavior (0 retries=100%, 1=70%, 2+=30%)

Status Classification

Status	Condition	Description
healthy	Score ≥ 80	Endpoint performing well
degraded	Score 50-79	Endpoint has issues but functional
unhealthy	Score < 50	Endpoint failing or critical issues

Input Example

Standard REST Endpoints

{
    "endpoints": [
        {
            "name": "Production API",
            "url": "https://api.example.com/health",
            "method": "GET",
            "expectedStatus": 200,
            "timeoutMs": 10000,
            "maxLatencyMs": 2000,
            "validateJsonKeys": ["status"]
        }
    ],
    "retryCount": 3,
    "notifyOnFailure": true
}

GenAI Endpoints

{
    "endpoints": [],
    "genaiEndpoints": [
        {
            "name": "OpenAI GPT-4",
            "provider": "openai",
            "model": "gpt-4",
            "apiKeyEnvVar": "OPENAI_API_KEY",
            "testPrompt": "Hi, respond with one word.",
            "maxTokens": 50,
            "timeoutMs": 30000
        },
        {
            "name": "Claude 3 Sonnet",
            "provider": "anthropic",
            "model": "claude-3-sonnet-20240229",
            "apiKeyEnvVar": "ANTHROPIC_API_KEY"
        },
        {
            "name": "Gemini Pro",
            "provider": "google",
            "model": "gemini-pro",
            "apiKeyEnvVar": "GOOGLE_AI_KEY"
        }
    ]
}

Output Example

{
    "summary": {
        "totalEndpoints": 3,
        "healthyEndpoints": 2,
        "degradedEndpoints": 1,
        "unhealthyEndpoints": 0,
        "averageLatencyMs": 450,
        "overallHealthScore": 85,
        "overallStatus": "degraded",
        "durationMs": 1523,
        "timestamp": "2024-01-15T10:00:00.000Z",
        "genai": {
            "totalGenAIEndpoints": 2,
            "totalTokensUsed": 150,
            "averageTTFTMs": null,
            "rateLimitedEndpoints": 0,
            "contentFilteredEndpoints": 0
        }
    },
    "results": [
        {
            "name": "OpenAI GPT-4",
            "url": "openai://gpt-4",
            "status": "healthy",
            "healthScore": 95,
            "latencyMs": 850,
            "isGenAI": true,
            "genaiMetrics": {
                "provider": "openai",
                "model": "gpt-4",
                "tokenUsage": {
                    "inputTokens": 10,
                    "outputTokens": 5,
                    "totalTokens": 15
                },
                "isRateLimited": false,
                "isContentFiltered": false,
                "responsePreview": "Hello!"
            }
        }
    ],
    "insights": {
        "recommendedActions": []
    }
}

GenAI Provider Configuration

Provider	Required Fields	Notes
openai	model, apiKeyEnvVar	Uses chat/completions API
anthropic	model, apiKeyEnvVar	Uses messages API
google	model, apiKeyEnvVar	Uses generateContent API
azure	model, apiKeyEnvVar, baseUrl	OpenAI-compatible format
custom	model, apiKeyEnvVar, baseUrl	For self-hosted or other LLMs

GenAI Endpoint Fields

Field	Type	Required	Default
name	String	✓	–
provider	String	✓	–
model	String	✓	–
apiKeyEnvVar	String	✓	–
baseUrl	String	*	Provider default
testPrompt	String		"Hi, respond with a single word."
maxTokens	Integer		50
timeoutMs	Integer		30000
streaming	Boolean		false

* Required for azure and custom providers

Insights & Recommendations

The Actor generates deterministic, rule-based recommendations:

Condition	Recommendation
Endpoint unreachable	Investigate - endpoint is unreachable
Wrong status code	Check - returning X instead of expected Y
Slow performance	Optimize - latency exceeds threshold
GenAI rate limited	Implement backoff or increase quota
Content filtered	Review test prompt or safety settings
High token usage	Reduce test prompt complexity

Environment Variables

Set API keys as environment variables:

OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_AI_KEY=AI...

Use {{VAR_NAME}} in REST endpoint headers:

{
    "headers": {
        "Authorization": "Bearer {{API_TOKEN}}"
    }
}

Programmatic Integration

const result = await apify.actor('your-actor').call(input);
const output = await result.dataset().listItems();
const report = output[0];

if (report.summary.overallStatus === 'unhealthy') {
    // Trigger alert
}

// Check GenAI health
if (report.summary.genai?.rateLimitedEndpoints > 0) {
    // Handle rate limiting
}

Limitations

• JSON key validation: top-level keys only
• Maximum timeout: 5 minutes per endpoint
• GenAI streaming TTFT: Coming in future release
• Retry delays capped at 30 seconds

Technical Specifications

Spec	Value
Max timeout	300,000 ms (5 min)
Max retry delay	30,000 ms
Min timeout	100 ms
Score range	0-100
GenAI max tokens	1-1000

Deployment

npm install
npm run build
apify push

Support

• Apify Documentation