Mattt

Hi @xmduhan. Thank you for opening this PR. Could you help me understand the use case here? Could you instead use the [`base_url` client parameter](https://github.com/replicate/replicate-python/blob/16c1520cf78414a290dca3e2cfe4079a50c8ed8c/replicate/client.py#L48) to specify a proxy?

@xmduhan Thanks for clarifying. I'm inclined to minimize the API surface area of the client itself, delegating additional behavior to httpx. Reading through [the docs](https://www.python-httpx.org/environment_variables/#http_proxy-https_proxy-all_proxy), I'm surprised that httpx didn't...

This would also help with a problem @zeke pointed out about how a model might change its inputs between versions.

Prediction inputs are now validated by the API.

@reckter Thanks for submitting this PR and apologies for the late response. Do we need a new argument for this? Or can we derive the formula name from the basename...

I created a Homebrew formula (see below), but got the following error when attempting to submit to Homebrew-core: ``` dependabot: * GitHub repository not notable enough (

The `gh` CLI does a good job of making itself available to different packaging systems, so we should follow their example: https://github.com/cli/cli/blob/trunk/docs/install_linux.md

My understanding is that this will cause incompatibilities with anyone using Pydantic 1, which I reckon is many if not most models. Any objections to porting this to use the...

Hi @tontan2545. Sorry to hear that you're having this problem in production. Other than an OOM, the only other cause that comes to mind is the server handling an explicit...

@tontan2545 [This FAQ](https://github.com/kubernetes/autoscaler/blob/master/cluster-autoscaler/FAQ.md) has some good background information about how autoscalers work in k8s generally. You can monitor autoscaler events with a command like: ```console kubectl logs -f deployment/cluster-autoscaler -n...