use 'native-tls' instead of rustls for reqwest

[SteiMi]

This aims to solve #36.

Instead of using 'rustls-tls' feature of reqwest, this uses 'native-tls'. This checks the certificates trusted by the operating system, which allows the user to easily trust custom self-signed certificates.

This built fine for me locally on Windows. Let's hope that the CI can build this successfully for all platforms.

Note: I also tried the feature 'rustls-tls-native-roots' which should theoretically also take certificates trusted by the OS into consideration, but I could not get it to trust my certificate.

[McPatate]

Is what you're looking to achieve not possible with rust-tls?

[FileMagic]

Before reading anything here just know I am a rust noob. I am just curious.

Is what you're looking to achieve not possible with rust-tls?

I think it is possible with rust-tls after some cursory glances at some docs.

I was just coming around to this for I saw that this repo is the one related to this issue here in llm-vscode. I think after reading this could instead be usage of rustls-native-certs instead of using native-tls.

Reading through the reqwests docs here tells me that the usage of native-tls will probs give us the best by platform import of tls certs trusted by the os.

However, what is the reasoning behind using rust-tls over native-tls? I don't see a reason to not use native-tls outside of the openssl compiling dependency?

TLDR: I think this MR is just to support using tls certs trusted by the host os, and I wonder if the reason they want to use native-tls is to just support tls certs by default on this host instead of having to disable tls which is what most people are doing.

[FileMagic]

Responding to myself here, I realized that you said in the #36 (comment) that you had issues with compiling using openssl.

So I think it would be more reasonable to switch to the usage of rustls-native-certs instead which exposes pub fn load_native_certs() -> Result<Vec<pki_types::CertificateDer<'static>>, std::io::Error>. I may write up a PR based on this here soon then to get a fix out for this issue.

[McPatate]

Closing in favor of #77