Production Deployment Best Practices Documentation

Mirascope / mirascope

LLM abstractions that aren't obstructions

https://docs.mirascope.io/

MIT License

680 stars 37 forks source link

Open willbakst opened 3 days ago

willbakst commented 3 days ago

We should include documentation specifically around best practices to deploy apps built using Mirascope to production.

This should cover things like error handling, retries, and more.

For anyone already deploying to production, feedback and input would be much appreciated here!

joekendal commented 3 days ago

with aws serverless constructs, we would be looking to have adapters for the stateful and stateless components:

chat history: dynamodb compute: lambda

willbakst commented 3 days ago

You should be able to use lambda to run the calls, and then pull chat history from dynamodb and insert it as a computed field.

The idea is that inside of your call you would pull the chat history from dynamo.

In the upcoming v1.3 release, this will be more clear when using prompt writing methods like messages shorthand.

Does this make sense?