Mattt

Hi @daneden. That's a great idea! We recently added support for [streaming prediction output](https://replicate.com/docs/streaming) to Replicate. #41 added an overload to `Client.createPrediction` that accepts a `stream` parameter. When set to...

@kristersj Sorry for the delay. This is now available in [0.24.0](https://github.com/replicate/replicate-swift/releases/tag/0.24.0).

Blocked by test cases not supporting different responses.

@aron Thanks for your feedback! I agree that delimiters would be a nice improvement. Curious to know what you think about the `scaffold` command running the example in the first...

We may also want to gracefully handle environments where Python is accessible from `python3` instead of `python`.

@jakedahn Not sure how widely used it is, but there is [Homebrew for Linux (and WSL2)](https://docs.brew.sh/Homebrew-on-Linux). Our [tap](https://github.com/replicate/homebrew-tap) CI tests for macOS and Linux, so I'm reasonably sure it works....

> That's nice if I only want the prediction output. But what if I want the whole prediction object? @zeke I'm pretty sure we don't document that anywhere, but `replicate.run`...

> Imagine coming upon this code without having read the docs. Would you understand that that callback was a progress event? Not that code golf version, no. But there are...

@zeke I think that gets to the underlying question: Do we want to make predictions more or less prominent in our system? I'd argue that a prediction a means to...

@onurkose FYI - `replicate.run` is a convenience method that combines `replicate.predictions.create` and `replicate.wait`. You can call `replicate.predictions.create` to create a prediction and not wait for it to finish, which is...