OpenRouter Model Availability

Overview

This skill covers model health monitoring, availability checking, and implementing automatic fallback chains for production reliability.

Prerequisites

• OpenRouter integration
• Multiple model options identified

Instructions

Follow these steps to implement this skill:

1. Verify Prerequisites: Ensure all prerequisites listed above are met
2. Review the Implementation: Study the code examples and patterns below
3. Adapt to Your Environment: Modify configuration values for your setup
4. Test the Integration: Run the verification steps to confirm functionality
5. Monitor in Production: Set up appropriate logging and monitoring

Overview

This skill covers model health monitoring, availability checking, and implementing automatic fallback chains for production reliability.

Prerequisites

• OpenRouter integration
• Multiple model options identified

Checking Model Status

List All Available Models

curl https://openrouter.ai/api/v1/models \
  -H "Authorization: Bearer $OPENROUTER_API_KEY"

Python Model Checker

import requests

def get_available_models(api_key: str) -> list:
    """Get all currently available models."""
    response = requests.get(
        "https://openrouter.ai/api/v1/models",
        headers={"Authorization": f"Bearer {api_key}"}
    )
    response.raise_for_status()
    return response.json()["data"]

def is_model_available(api_key: str, model_id: str) -> bool:
    """Check if specific model is available."""
    models = get_available_models(api_key)
    return any(m["id"] == model_id for m in models)

# Usage
if is_model_available(api_key, "openai/gpt-4-turbo"):
    print("Model is available")

TypeScript Model Checker

interface ModelInfo {
  id: string;
  name: string;
  context_length: number;
  pricing: {
    prompt: string;
    completion: string;
  };
}

async function getAvailableModels(apiKey: string): Promise<ModelInfo[]> {
  const response = await fetch('https://openrouter.ai/api/v1/models', {
    headers: { Authorization: `Bearer ${apiKey}` },
  });
  const data = await response.json();
  return data.data;
}

async function isModelAvailable(apiKey: string, modelId: string): Promise<boolean> {
  const models = await getAvailableModels(apiKey);
  return models.some((m) => m.id === modelId);
}

Model Health Monitoring

Health Check Function

import time
from typing import Optional

def check_model_health(
    model_id: str,
    timeout: float = 10.0
) -> dict:
    """Check if model responds correctly."""
    start = time.time()

    try:
        response = client.chat.completions.create(
            model=model_id,
            messages=[{"role": "user", "content": "Say 'ok'"}],
            max_tokens=5,
            timeout=timeout
        )

        latency = time.time() - start
        return {
            "model": model_id,
            "status": "healthy",
            "latency_ms": latency * 1000,
            "response": response.choices[0].message.content
        }

    except Exception as e:
        latency = time.time() - start
        return {
            "model": model_id,
            "status": "unhealthy",
            "latency_ms": latency * 1000,
            "error": str(e)
        }

# Check multiple models
models_to_check = [
    "openai/gpt-4-turbo",
    "anthropic/claude-3.5-sonnet",
    "meta-llama/llama-3.1-70b-instruct"
]

for model in models_to_check:
    health = check_model_health(model)
    print(f"{model}: {health['status']} ({health['latency_ms']:.0f}ms)")

Continuous Monitoring

import asyncio
from datetime import datetime

class ModelMonitor:
    def __init__(self, models: list, check_interval: int = 60):
        self.models = models
        self.check_interval = check_interval
        self.status_history = {m: [] for m in models}

    async def check_all(self):
        """Check all models once."""
        results = {}
        for model in self.models:
            health = check_model_health(model)
            results[model] = health

            self.status_history[model].append({
                "timestamp": datetime.now().isoformat(),
                "status": health["status"],
                "latency": health.get("latency_ms", 0)
            })

            # Keep last 100 entries
            self.status_history[model] = self.status_history[model][-100:]

        return results

    def get_uptime(self, model: str) -> float:
        """Calculate uptime percentage for model."""
        history = self.status_history.get(model, [])
        if not history:
            return 0.0
        healthy = sum(1 for h in history if h["status"] == "healthy")
        return healthy / len(history) * 100

    async def run(self):
        """Continuous monitoring loop."""
        while True:
            results = await self.check_all()
            print(f"\n[{datetime.now().isoformat()}] Model Status:")
            for model, health in results.items():
                uptime = self.get_uptime(model)
                print(f"  {model}: {health['status']} (uptime: {uptime:.1f}%)")
            await asyncio.sleep(self.check_interval)

# Usage
monitor = ModelMonitor([
    "openai/gpt-4-turbo",
    "anthropic/claude-3.5-sonnet"
])
# asyncio.run(monitor.run())

Handling Model Unavailability

Fallback Chain

class ModelFallback:
    def __init__(self, models: list):
        self.models = models
        self.disabled = set()
        self.disable_until = {}

    def disable_model(self, model: str, duration: float = 300):
        """Temporarily disable a model."""
        self.disabled.add(model)
        self.disable_until[model] = time.time() + duration

    def get_available_models(self) -> list:
        """Get currently available models."""
        now = time.time()
        # Re-enable models past their disable duration
        for model in list(self.disabled):
            if self.disable_until.get(model, 0) < now:
                self.disabled.remove(model)

        return [m for m in self.models if m not in self.disabled]

    def chat(self, prompt: str, **kwargs):
        """Try models in order until one works."""
        available = self.get_available_models()

        if not available:
            raise Exception("All models unavailable")

        for model in available:
            try:
                return client.chat.completions.create(
                    model=model,
                    messages=[{"role": "user", "content": prompt}],
                    **kwargs
                )
            except Exception as e:
                if "unavailable" in str(e).lower():
                    self.disable_model(model)
                    continue
                raise

        raise Exception("All models failed")

# Usage
fallback = ModelFallback([
    "anthropic/claude-3.5-sonnet",
    "openai/gpt-4-turbo",
    "meta-llama/llama-3.1-70b-instruct"
])

Model Groups

MODEL_GROUPS = {
    "coding": [
        "anthropic/claude-3.5-sonnet",
        "openai/gpt-4-turbo",
        "deepseek/deepseek-coder"
    ],
    "fast": [
        "anthropic/claude-3-haiku",
        "openai/gpt-3.5-turbo",
        "meta-llama/llama-3.1-8b-instruct"
    ],
    "cheap": [
        "meta-llama/llama-3.1-8b-instruct",
        "mistralai/mistral-7b-instruct"
    ]
}

def get_working_model(group: str) -> Optional[str]:
    """Find first working model in group."""
    models = MODEL_GROUPS.get(group, [])

    for model in models:
        health = check_model_health(model, timeout=5.0)
        if health["status"] == "healthy":
            return model

    return None

Provider Status

Check Provider Health

PROVIDERS = {
    "openai": ["openai/gpt-4-turbo", "openai/gpt-3.5-turbo"],
    "anthropic": ["anthropic/claude-3.5-sonnet", "anthropic/claude-3-haiku"],
    "meta": ["meta-llama/llama-3.1-70b-instruct", "meta-llama/llama-3.1-8b-instruct"],
}

def check_provider_health(provider: str) -> dict:
    """Check health of all models from a provider."""
    models = PROVIDERS.get(provider, [])
    results = {"provider": provider, "models": {}}

    for model in models:
        health = check_model_health(model)
        results["models"][model] = health["status"]

    healthy = sum(1 for s in results["models"].values() if s == "healthy")
    results["healthy_count"] = healthy
    results["total_count"] = len(models)
    results["status"] = "healthy" if healthy == len(models) else "degraded" if healthy > 0 else "down"

    return results

Status Dashboard

Simple Status Output

def print_status_dashboard():
    """Print formatted status dashboard."""
    print("=" * 60)
    print("OpenRouter Model Status Dashboard")
    print("=" * 60)

    models = get_available_models(api_key)

    # Group by provider
    by_provider = {}
    for model in models:
        provider = model["id"].split("/")[0]
        if provider not in by_provider:
            by_provider[provider] = []
        by_provider[provider].append(model)

    for provider, provider_models in sorted(by_provider.items()):
        print(f"\n{provider.upper()}:")
        for model in provider_models[:5]:  # Show first 5
            ctx = model.get("context_length", "?")
            prompt_price = model.get("pricing", {}).get("prompt", "?")
            print(f"  ✓ {model['id']} (ctx: {ctx}, ${prompt_price}/token)")
        if len(provider_models) > 5:
            print(f"  ... and {len(provider_models) - 5} more")

print_status_dashboard()

Model Comparison Table

def compare_models(model_ids: list):
    """Compare models side by side."""
    models = get_available_models(api_key)
    model_map = {m["id"]: m for m in models}

    print(f"{'Model':<40} {'Context':<10} {'Prompt $/M':<12} {'Completion $/M'}")
    print("-" * 80)

    for model_id in model_ids:
        if model_id in model_map:
            m = model_map[model_id]
            ctx = m.get("context_length", "N/A")
            prompt = float(m["pricing"]["prompt"]) * 1_000_000
            completion = float(m["pricing"]["completion"]) * 1_000_000
            print(f"{model_id:<40} {ctx:<10} ${prompt:<11.2f} ${completion:.2f}")
        else:
            print(f"{model_id:<40} NOT AVAILABLE")

compare_models([
    "openai/gpt-4-turbo",
    "anthropic/claude-3.5-sonnet",
    "anthropic/claude-3-haiku"
])

Output

Successful execution produces:

• Working OpenRouter integration
• Verified API connectivity
• Example responses demonstrating functionality

Error Handling

Common errors and solutions:

1. 401 Unauthorized: Check API key format (must start with sk-or-)
2. 429 Rate Limited: Implement exponential backoff
3. 500 Server Error: Retry with backoff, check OpenRouter status page
4. Model Not Found: Verify model ID includes provider prefix

Examples

See code examples in sections above for complete, runnable implementations.

Resources

• OpenRouter Documentation
• OpenRouter Models
• OpenRouter API Reference
• OpenRouter Status

Output

Successful execution produces:

• Working OpenRouter integration
• Verified API connectivity
• Example responses demonstrating functionality

Error Handling

Common errors and solutions:

1. 401 Unauthorized: Check API key format (must start with sk-or-)
2. 429 Rate Limited: Implement exponential backoff
3. 500 Server Error: Retry with backoff, check OpenRouter status page
4. Model Not Found: Verify model ID includes provider prefix

Examples

See code examples in sections above for complete, runnable implementations.

Resources

• OpenRouter Documentation
• OpenRouter Models
• OpenRouter API Reference
• OpenRouter Status