Skip to main content
When using Universal AI / expert models, there is no built-in fallback mechanism in the model string (unlike the LLM smart routing). Instead, you implement fallback logic in your application code.

​
What is Fallback?

Fallback means automatically retrying a request with an alternative provider when the primary provider fails. This improves reliability by ensuring your application continues working even if one provider experiences downtime or errors.

​
Implementation Pattern

The approach is straightforward:
  1. Send a request to your primary provider
  2. If it fails (network error, provider outage, rate limit, etc.), catch the error
  3. Retry the same request with a fallback provider
import requests

url = "https://api.edenai.run/v3/universal-ai"
headers = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json"
}

text_to_moderate = "Some text to check for harmful content."

# Define primary and fallback models
primary_model = "text/moderation/openai"
fallback_model = "text/moderation/google"

def call_universal_ai(model, text):
    response = requests.post(url, headers=headers, json={
        "model": model,
        "input": {"text": text}
    })
    response.raise_for_status()
    return response.json()

# Try primary, fall back if it fails
try:
    result = call_universal_ai(primary_model, text_to_moderate)
    print("Primary provider succeeded")
except requests.exceptions.RequestException as e:
    print(f"Primary provider failed: {e}. Trying fallback...")
    result = call_universal_ai(fallback_model, text_to_moderate)
    print("Fallback provider succeeded")

print(result)

​
Multiple Fallbacks

You can chain several fallback providers for even greater resilience:
Python
import requests

url = "https://api.edenai.run/v3/universal-ai"
headers = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json"
}

models = [
    "text/moderation/openai",
    "text/moderation/google",
    "text/moderation/microsoft"
]

result = None
for model in models:
    try:
        response = requests.post(url, headers=headers, json={
            "model": model,
            "input": {"text": "Text to moderate"}
        })
        response.raise_for_status()
        result = response.json()
        print(f"Success with {model}")
        break
    except requests.exceptions.RequestException as e:
        print(f"{model} failed: {e}")

if result is None:
    print("All providers failed")

​
Best Practices

Choose fallback providers that support the same feature and subfeature. For example, if your primary is text/moderation/openai, your fallback should also be a text/moderation/* provider. Use the Listing Models endpoint to discover which providers support a given feature.
  • Order by preference — place your most reliable or cost-effective provider first.
  • Log fallback events — track when fallbacks occur so you can identify recurring provider issues.
  • Set timeouts — add request timeouts so a slow provider triggers the fallback quickly rather than hanging.
  • Handle all providers failing — always have a final error path if every provider is unavailable.