when i was building silkworm, unability of uvloop to support Windows was a problem for me, so i added winloop

btw, they seem to be included into uvloop: https://github.com/Vizonex/Winloop/commit/317873329143e8cf82...

added support for Windows in v0.1.3

Also, it supports Free-threaded Python (`PYTHON_GIL=0` env-var).

My little test that extracts title's from webpages (spider https://github.com/BitingSnakes/silkworm/blob/main/examples/...):

- RPS with GIL = ~174 - RPS without GIL: ~242

I've built https://github.com/RustedBytes/scraper-rs to parse HTML using Rust with CSS selectors and XPath expressions. This wrapper can be useful for others as well.

Here's demos:

- https://huggingface.co/spaces/Yehor/en-uk-translator

- https://huggingface.co/spaces/Yehor/uk-en-translator

Are you using a local Whisper? If yes, what do you use for inference, candle/ort?

Not local. Inference is the only part not written in Rust so far.

I am using Replicate to run docker images with a pipeline based on faster-whipser, VAD, pyannote and a custom LLM enhancement flow.

Thanks for sharing candle/ort. Interesting to see the WASM in-browser opportunities

Not easy it sounds. Generated file has JSONL and each has summary or advisory lines. My script just processes them to a Markdown in Go.

Actually, it's the same. As I understand they use one database.

Just added the quantized version of the model

I agree with this position. The code and the API spec must be separated. But the project is not meant to write APIs with spec, the project uses Python to create API spec.