This is a Cog model for testing failure modes in applications that consume Cog models.

This model acts as a “chaos monkey”, deliberately introducing controlled failures and unexpected behaviors to help test the resilience of systems that integrate with Cog models. It’s designed to simulate different types of responses that might occur in production environments, allowing developers to ensure their applications handle these scenarios gracefully.

Usage

You can run this model on the Replicate website, or via API, or locally with the Cog CLI:

Basic Usage:

cog predict -i outcome=success

With delay:

cog predict -i outcome=success -i sleep=5

Testing error handling:

cog predict -i outcome=error

Testing bad data handling:

cog predict -i outcome=bad_return_type

Testing random data handling:

cog predict -i outcome=random_data