bjabrboe1984

I wrote about building production LLM workflows in JavaScript and created a complete working example.

The example demonstrates: - Multi-step workflow state management - OpenAI API integration with automatic retries and exponential backoff - Caching to reduce costs - GitHub webhook integration with signature verification - Serverless deployment (no infrastructure to manage)

This sits between lightweight prompt libraries (LangChain) and heavyweight orchestrators (Airflow). It's for developers who need state persistence, retries, and scheduling without spinning up new infrastructure.

Live example code: https://github.com/RestDB/codehooks-io-examples/tree/main/ll...

Blog post with details: https://codehooks.io/blog/building-llm-workflows-javascript

The example is fully functional - you can deploy it in 5 minutes with `npm install` and `coho deploy`.

Thanks for taking the time to answer!

The main difference is that codehooks.io embeds a document database, a key-value store, a job scheduler and a worker queue without any config/setup. We want it to be easy to create a complete backend using codehooks.io in the shortest possible time.

Another difference is the speed and simplicity of deployment. Deployment typically takes 1 second or less. This increases developer productivity.

Our goal is that the developer can focus 100% on the application logic and 0% on the setup.

We have considered open sourcing it and we might do that at a later stage.

Create JS (or TS) hooks (functions), use two types of NoSQL (document, key/value), message queues, cron jobs. Deploy! No config or complex setup like K8.

Fieldbook.com