There are many use cases where we're not trying to build real-time ChatGPT-like experiences, so the underlying technology should support asynchronous mechanisms such as callback to enable applications to work more efficiently.
Examples include:
- Processing or generating large text documents, images, or video.
- Batch processing inputs.
- Piping the output from an AI process to another service with no synchronous response required.
In these situations, you don't want to hold a connection open to an API and await a response. Instead, you want to receive a callback when the process has completed.
All of these scenarios represent the need to manage asynchronous messaging.
- Replicate: webhook callbacks for persisting prediction data and files, sending notifications when long-running predictions finish, and creating model pipelines.
- AssemblyAI: get notified when your transcripts are ready.
- Deepgram: return transcripts to a callback URL sent to Deepgram's API.
- Hookdeck is an event gateway platform that supports ingesting events from platforms, localhost development, and routing events to other services.
- Resend: Email API that developers love offering webhook callbacks for various email events.
- Twilio
- Programmable SMS: Webhooks for inbound SMS. Note that to support asynchronous workflows, you'll need to follow the "Messaging without responding" workflow.
- Programmable Voice: notifications about call status progress and trigger call workflows via webhook callbacks.
Generate images with Replicate, receive the images on localhost, and in parallel send an email with an image with Resend.
