Maybe I'm missing something, but it seems like there is no obvious place for me to just copy a flake-url from. I have to piece the output together myself based on my knowledge of flake-URLs, but especially beginners won't have that knowledge yet.

It would be nice if there was a button next to the outputs, or maybe a list of URLs in the details page of every output, that I could quickly copy into inputs.$whatever.url, nix shell, nix run etc.

For privacy/security I allowlist sites to enable JavaScript. When I visited the site, I got a blank white page with no information.

I would really appreciate being able to specify <flakestry-url>/1.2.* as an input, so that an update fetches the latest tag for that range.
I assume this would eventually be possible with general flake versioning mechanisms, but it would be great if flakestry could provide this ad-hoc in the meantime!

Support once NixOS/nix#8892 is merged

Current logo is generated using https://www.midjourney.com/ as a snowman.

Suggestions welcome!

ResponseValidationError: 1 validation errors:
{'loc': ('response',), 'msg': 'none is not an allowed value', 'type': 'type_error.none.not_allowed'}

Basically like the on flakehub has

Requires a few things to happen:

• auth
• /publish endpoint needs to accept tarball as POST body
• async job to process tarball (evaluate flake, render readme, etc)

What needs to happen for this to replace https://search.nixos.org/flakes?

Maybe we should ping the static registry owners to push flakes into it.

First and foremost, I'd like to thank everyone who contributed to this project, which has been an excellent contribution to the Nix ecosystem. Inspired by @domenkozar work on devenv, I want to propose rewriting the backend in Rust. While some of the primary contributors may already be familiar with this language, I would want to summarize and highlight some of the features that this project would benefit the most from:

• Fearless concurrency - Rust maintains memory safety, which is particularly beneficial in the context of concurrent programming because it prevents the majority of typical issues such as data races. More specifically, the publish endpoint may benefit from these because tasks like inserting a new record and indexing could be done concurrently.
• Error handling - Rust distinguishes between recoverable and unrecoverable errors, making it easier to reason about all the possible edge cases of functions, as errors are part of the type signature Result<T, E> and need to be handled explicitly.

After some little research, I found that there are Rust clients for all of the services the application relies on:

• Open Search - There is an official client.
• GitHub - While there are a few alternatives here, it is also possible to call the GitHub API directly using a simple HTTP client like hyper or reqwest, which is quite similar to what current Python code does.
• PostgreSQL - While there are a few native libraries to communicate directly with PostgreSQL, there are also database-agnostic options like SQLx that check queries at compile time.

Regarding web frameworks, axum appears to be one of the most popular based on overall performance. Before getting into more technical details, I wanted to get some feedback about this proposal, as I understand that rewriting a piece of software in a new language always involves some risk and has an impact on maintainers. Nonetheless, I'd like to put this on the table in order to start a conversation and, hopefully, answer all of the questions that arise.

Note: If this proposal gets approved, I would be pleased to submit a PR.

E.g. see the "SSH Access Module" in https://staging.flakestry.dev/flake/github/infinisil/nixus/0.1.

It seems like markdown rendering is done using https://github.com/elm-explorations/markdown/tree/1.0.0, which is based on https://github.com/markedjs/marked, which does have a known extension to handle this: https://www.npmjs.com/package/marked-base-url, though it doesn't seem directly supported by the markdown Elm package

Hello! Nice to see another attempt for a flake registry :)

From what I read in on the publish page, flakestry.dev currently only supports flakes working with tags, is that correct? We would like to publish Nyx here, but we don't make use of tags and mostly have daily commits.

This package gives a decoder error: https://flakestry.dev/flake/github/infinisil/nixus/0.1.3

The Json.Decode.oneOf at json.packages failed in the following 2 ways:



(1) Problem with the value at json['aarch64-linux'].example:
    
        {}
    
    Expecting an OBJECT with a field named `type`



(2) Problem with the given value:
    
    {
            "x86_64-linux": {
                "example": {
                    "name": "deploy",
                    "type": "derivation"
                }
            },
            "aarch64-linux": {
                "example": {}
            },
            "x86_64-darwin": {
                "example": {}
            },
            "aarch64-darwin": {
                "example": {}
            }
        }
    
    Expecting null

Render Github markdown syntax

> [!IMPORTANT]
> Render it

> [!NOTE] 
> Render it

> [!WARNING] 
> Render it

Example

Example of how it is done currently.

First off, this is cool, thanks for building it!

I have a quick UX suggestion. Because some packages have extensive semver histories, it would be helpful if the versions were reverse-sorted, with the latest at the top. This small change could make finding the newest versions much quicker and more intuitive.

Do nix flake metadata --json and nix flake show --json representations.

These are already stored in db, we need to write an Elm parser for it and display the outputs.

https://testdriven.io/blog/fastapi-sqlmodel/

https://package.elm-lang.org/packages/ryannhg/date-format/latest/DateFormat-Relative

Despite the fact that the status badge renders fine on flakestry.dev, it appears that the badge from the clipboard is broken. Here is an example copied directly from the clipboard:

Currently we don't filter out older version from search.

Going to http://staging.flakestry.dev/ just hangs, it doesn't seem to be listened to. There should be a redirect to https://staging.flakestry.dev/.

I don't fully get fly.toml, but it seems like the force_https flag might fix this: https://fly.io/docs/reference/configuration/#services-ports