A library for quickly building bluesky feed generators.
Primarily uses, warp, atrium api, and jetstream-oxide to greatly simplify the process of building a bluesky feed generator.
Create a .env file with the following variables:
PUBLISHER_DID
Your DID.
This can be a little hard to track down - you can use this utility to check your DID
To run the my_did utility - clone this repo & run this command inside the crate directory
cargo run --bin my_did --handle <your handle> --app-password <app password>
PUBLISHER_DID="..."
FEED_GENERATOR_HOSTNAME
The host name for your feed generator.
(In the URL https://github.com/cyypherus/skyfeed
the host name is github.com
)
You can develop your feed locally without setting this to a real value. However, when publishing, this value must be a domain that:
- Points to your service.
- Is secured with SSL (HTTPS).
- Is accessible on the public internet.
FEED_GENERATOR_HOSTNAME="..."
Once published, or while testing, your feed will be served at http://<host name>/xrpc/app.bsky.feed.getFeedSkeleton?feed=<feed name>
.
Documentation on additional query parameters is available here.
Let's build a simple feed generator about cats.
Note: In a real implementation storage should be implemented with a database such as sqlite for more efficient queries & persistent data. See the sqlite example
Your feed handler is responsible for storing and managing firehose input. For the sake of simplicity, we'll just use Vec and HashMap to manage posts and likes.
#[derive(Clone)]
struct MyFeedHandler {
posts: Arc<Mutex<Vec<MyPost>>>,
}
#[derive(Debug, Clone)]
struct MyPost {
post: Post,
likes: HashSet<Uri>,
}
impl FeedHandler for MyFeedHandler {
async fn insert_post(&mut self, post: Post) {
if post.text.to_lowercase().contains(" cat ") {
info!("Storing {post:?}");
const MAX_POSTS: usize = 100;
let mut posts = self.posts.lock().await;
posts.push(MyPost {
post,
likes: HashSet::new(),
});
if posts.len() > MAX_POSTS {
posts.remove(0);
}
}
}
async fn delete_post(&mut self, uri: Uri) {
self.posts
.lock()
.await
.retain(|post_with_likes| post_with_likes.post.uri != uri);
}
async fn like_post(&mut self, like_uri: Uri, liked_post_uri: Uri) {
if let Some(post_with_likes) = self
.posts
.lock()
.await
.iter_mut()
.find(|p| p.post.uri == liked_post_uri)
{
post_with_likes.likes.insert(like_uri);
}
}
async fn delete_like(&mut self, like_uri: Uri) {
let mut posts = self.posts.lock().await;
for post_with_likes in posts.iter_mut() {
post_with_likes.likes.remove(&like_uri);
}
}
async fn serve_feed(&self, request: Request) -> FeedResult {
info!("Serving {request:?}");
let posts = self.posts.lock().await;
// Parse the cursor from the request
let start_index = if let Some(cursor) = &request.cursor {
cursor.parse::<usize>().unwrap_or(0)
} else {
0
};
let posts_per_page = 5;
// Sort posts by likes
let mut sorted_posts: Vec<_> = posts.iter().collect();
sorted_posts.sort_by(|a, b| b.likes.len().cmp(&a.likes.len()));
// Paginate posts
let page_posts: Vec<_> = sorted_posts
.into_iter()
.skip(start_index)
.take(posts_per_page)
.cloned()
.collect();
// Calculate the next cursor
let next_cursor = if start_index + posts_per_page < posts.len() {
Some((start_index + posts_per_page).to_string())
} else {
None
};
FeedResult {
cursor: next_cursor,
feed: page_posts
.into_iter()
.map(|post_with_likes| post_with_likes.post.uri.clone())
.collect(),
}
}
}
We'll need to use Arc<Mutex<FeedHandler>>
to enable concurrent shared access.
struct MyFeed {
handler: MyFeedHandler,
}
impl Feed<MyFeedHandler> for MyFeed {
fn handler(&mut self) -> MyFeedHandler {
self.handler.clone()
}
}
Now we can create an instance of our Feed
and start it on a local address.
#[tokio::main]
async fn main() {
let mut feed = MyFeed {
handler: MyFeedHandler {
posts: Arc::new(Mutex::new(Vec::new())),
},
};
feed.start("Cats", ([0, 0, 0, 0], 3030)).await
}
This repo also contains publish (and unpublish) utilities for managing your feed's publicity.
To run these, clone this repo & run this command inside the crate directory
cargo run --bin publish
If you'd like to verify your feed server's endpoints locally before you publish, you can also use the verify utility.