Skip to main content
The Redis queue API is no longer supported and has been removed from Cog.

​
Overview

The Redis queue worker functionality that was previously available in Cog has been deprecated and removed. This feature allowed models to process predictions asynchronously using Redis as a message queue backend.

​
Migration

If you were using the Redis queue worker in previous versions of Cog, you have several alternatives:

​
Use HTTP Endpoints

The standard HTTP server mode provides a synchronous API for running predictions: 
cog predict
This runs predictions through Cog’s HTTP server, which handles requests synchronously and returns results directly.

​
External Queue Solutions

For asynchronous processing, you can integrate Cog with external queue systems:
  • Celery - Python-based distributed task queue
  • RQ (Redis Queue) - Simple Python job queue
  • BullMQ - Redis-based queue for Node.js applications
  • Cloud-native queues - AWS SQS, Google Cloud Tasks, Azure Queue Storage
When building your own queue integration, use Cog’s HTTP API to invoke predictions. This gives you full control over queue behavior, retry logic, and error handling.

​
Deployment Options

For production deployments with built-in queue management, consider:
  • Replicate - Fully managed platform for running machine learning models with automatic scaling and queue management
  • Custom infrastructure - Build your own queue system around Cog’s HTTP server using container orchestration platforms like Kubernetes

​
Questions?

If you need help migrating away from the Redis queue worker or implementing an alternative solution, please refer to the main Cog documentation or open an issue in the GitHub repository.

Build docs developers (and LLMs) love

Get started for freeTalk to us