Atmosfeed

Create custom Bluesky feeds with WebAssembly and Scale.

Overview

🚧 This project is a work-in-progress! Instructions will be added as soon as it is usable. 🚧

Installation

Atmosfeed is available to the public and can be used by opening it in a browser:

If you prefer to self-host, see Contributing; static binaries for the manager and worker, a .tar.gz archive for the frontend and an OCI image for containerization are also available on GitHub releases and GitHub container registry respectively.

In addition to this publicly hosted infrastructure, a client CLI is also provided in the form of static binaries in order to make feed deployment and management from your local system easier.

On Linux, you can install them like so:

curl -L -o /tmp/atmosfeed-client "https://github.com/pojntfx/atmosfeed/releases/latest/download/atmosfeed-client.linux-$(uname -m)"
sudo install /tmp/atmosfeed-client /usr/local/bin

On macOS, you can use the following:

curl -L -o /tmp/atmosfeed-client "https://github.com/pojntfx/atmosfeed/releases/latest/download/atmosfeed-client.darwin-$(uname -m)"
sudo install /tmp/atmosfeed-client /usr/local/bin

On Windows, the following should work (using PowerShell as administrator):

Invoke-WebRequest https://github.com/pojntfx/atmosfeed/releases/latest/download/atmosfeed-client.windows-x86_64.exe -OutFile \Windows\System32\atmosfeed-client.exe

You can find binaries for more operating systems and architectures on GitHub releases.

Usage

1. Accessing an Atmosfeed Server

A public Atmosfeed server is provided to you at https://manager.atmosfeed.p8.lu; if you prefer to host your own, see Contributing for deployment instructions.

2. Creating a Feed Classifier

Atmosfeed allows creating custom "classifiers" in order to create feeds. A classifier is a Scale function written in one of the supported languages (currently Go, Rust and TypeScript) that is compiled to WebAssembly .scale function. This classifier is then run for each new skeet posted to Bluesky, and returns a weight that determines whether the skeet will show up in the custom feed and at which position (the higher the weight, the higher the skeet's position in the feed). This allows creating feeds that are about a topic, in a specific language, from specific people, contain specific words etc., and also allows you to customize the feed order to be chronological, based on like count or any other metric.

Prefer to start with an example instead or don't have access to a terminal? Download one of the example feeds from their download site and jump to either Testing a Classifier Locally or Pushing the Classifier.

To create a classifier, start by installing the Scale CLI and then run the following to create a new classifier based on the felicitaspojtinger/classifier:latest signature:

scale function new trending:latest -s felicitaspojtinger/classifier:latest -l go -d trending

We've named this classifier trending and will use it to build a feed that returns the most popular posts in English; we've chosen Go as the language of choice, but you could also have switched -l go to -l rust to use Rust or -l ts to use TypeScript.

The generated main.go will look like this:

package trending

import (
	"signature"
)

func Scale(ctx *signature.Context) (*signature.Context, error) {
	return signature.Next(ctx)
}

First, let's change this feed to return only English posts by returning a positive weight only to posts with the en language element:

func Scale(ctx *signature.Context) (*signature.Context, error) {
	if len(ctx.Post.Langs) == 1 && ctx.Post.Langs[0] == "en" {
		ctx.Weight = 1
	} else {
		ctx.Weight = -1
	}

	return signature.Next(ctx)
}

Since this will return the same weight for all posts, this feed would return posts based on insertion order. To instead return them based on their creation date, use ctx.Weight = ctx.Post.CreatedAt; to make this a trending feed, simply weight them by likes:

func Scale(ctx *signature.Context) (*signature.Context, error) {
	if len(ctx.Post.Langs) == 1 && ctx.Post.Langs[0] == "en" {
		ctx.Weight = ctx.Post.Likes
	} else {
		ctx.Weight = -1
	}

	return signature.Next(ctx)
}

3. Testing a Classifier Locally

First, build the classifier to WebAssembly using the Scale CLI:

scale function build --release -d trending

Then, export the classifier to a .scale file:

scale function export local/trending:latest trending/out

Using the Atmosfeed CLI, you can now connect to the Bluesky classifier and run the classifier for each new post on your local system:

atmosfeed-client dev --feed-classifier trending/out/local-trending-latest.scale

As posts are being created, the classifier's weight (which will determine the post's order) as well as the URL of each post are logged to your terminal:

2023/11/25 22:26:16 Connected to BGS https://bsky.network
5 https://bsky.app/profile/did:plc:lyarbqrrgmm2zjcclcnpxfpd/post/3kf24w3mtuo2b {did:plc:lyarbqrrgmm2zjcclcnpxfpd 3kf24w3mtuo2b I remain amazed that Fallout 3 let you use pickpocket to put live grenades in people's pockets and walk away for them to die horribly. [en] 5 0 true}
7 https://bsky.app/profile/did:plc:govebcmu5zv67cpy3qdchgg4/post/3kf24w3io3n2a {did:plc:govebcmu5zv67cpy3qdchgg4 3kf24w3io3n2a ALSO thoroughly enjoyed the title and credits sequence revisiting the Matt Smith era clouds motif. Looked tremendous. [en] 7 0 true}
20 https://bsky.app/profile/did:plc:2i3rr5wflkbtfstshwmvbn2i/post/3kf24w3u2cm2m {did:plc:2i3rr5wflkbtfstshwmvbn2i 3kf24w3u2cm2m I often have to finish the edges with a knife. One of these days I'll have to start with a knife too. [en] 20 0 true}
5 https://bsky.app/profile/did:plc:usl5fc2l73alvigqdel377ss/post/3kf24w3ns3v2f {did:plc:usl5fc2l73alvigqdel377ss 3kf24w3ns3v2f On the day I first committed myself to a life of Zen practice. 'Zen Sickness', Zen Master Hakuin.
t.co/yIKUfttd0s [en] 5 0 false}

4. Pushing the Classifier

After building and testing a classifier locally, you can continue by pushing it to the Atmosfeed server. To do so, you can either use the Atmosfeed CLI (see Command Line Arguments for more information):

atmosfeed-client apply --feed-rkey trending --feed-classifier trending/out/local-trending-latest.scale

Or by visiting the Atmosfeed UI and using the "Create a new feed" wizard after signing in with your Bluesky account:

In the wizard, you can then select the rkey of your feed (this is what Bluesky uses to uniquely identify the feeds of a user; you'll be able to set the display name at a later point); it is also possible to specify a pinned posts, which will then always show up at the top of your feed:

5. Publishing a Feed

Once you have uploaded the a feed classifier to Atmosfeed, you can publish your feed to Bluesky. One option to do this is to use the Atmosfeed CLI like so:

atmosfeed-client publish --feed-rkey trending --feed-name 'Trending' --feed-description 'Most popular trending posts on Bluesky'

Alternatively, you can do this using the Atmosfeed UI by clicking the "Finalize" button for your uploaded feed:

This will open up the feed finalization/publishing wizard, where it is possible to set the display name and a description:

After having published a feed, it will show up in your list of published feeds:

It is also possible to retrieve this list by using the Atmosfeed CLI:

atmosfeed-client list

Which will return it in YAML format:

- rkey: trending
  pinneddid: ""
  pinnedrkey: ""

The feed should also show up in the Bluesky UI under Your Account → Feeds:

6. Updating a Feed and Classifier

It is possible to update a feed and the associated classifier after pushing/publishing. To update a feed classifier or the pinned post using the Atmosfeed CLI, simply run apply again:

atmosfeed-client apply --feed-rkey trending --feed-classifier trending/out/local-trending-latest.scale

Please note that empty values for --pinned-feed-did and --pinned-feed-rkey are ignored in order to allow updating the classifier in isolation; if you want to set them to empty values, pass --clear-pinned.

To update a published feed's values, you can simply publish it again:

atmosfeed-client publish --feed-rkey trending --feed-name 'Trending' --feed-description 'Most popular trending posts on Bluesky'

In order to update a pushed feed or a published feed using the Atmosfeed UI, you can simply use the edit wizard:

7. Deleting a Feed and Classifier

Both the Atmosfeed CLI and UI support deleting feed classifiers from Atmosfeed and unpublishing feeds from Bluesky.

To unpublish a feed, but keep the classifier, you can use the unpublish command:

atmosfeed-client unpublish --feed-rkey trending

To then delete the classifier, use the delete command:

atmosfeed-client delete --feed-rkey trending

Similarly so, you can use the Atmosfeed UI to unpublish a feed by using the unpublish option in the feed list:

Once unpublished, you can delete the feed's classifier by using the delete option in the feed list:

🚀 That's it! We can't wait to see what you're going to build with Atmosfeed. Be sure to take a look at the reference and example feeds for more information.

Reference

Command Line Arguments

Server CLI

$ atmosfeed-server --help
Create custom Bluesky feeds with WebAssembly and Scale.
Find more information at:
https://github.com/pojntfx/atmosfeed

Usage:
  atmosfeed-server [command]

Available Commands:
  completion  Generate the autocompletion script for the specified shell
  help        Help about any command
  manager     Start an Atmosfeed manager
  worker      Start an Atmosfeed worker

Flags:
  -h, --help                  help for atmosfeed-server
      --postgres-url string   PostgreSQL URL (default "postgresql://postgres@localhost:5432/atmosfeed?sslmode=disable")
      --redis-url string      Redis URL (default "redis://localhost:6379/0")
      --s3-url string         S3 URL (default "http://minioadmin:minioadmin@localhost:9000?bucket=atmosfeed")
      --verbose               Whether to enable verbose logging

Use "atmosfeed-server [command] --help" for more information about a command.

Expand subcommand reference

Manager

$ atmosfeed-server manager --help
Start an Atmosfeed manager

Usage:
  atmosfeed-server manager [flags]

Aliases:
  manager, m

Flags:
      --bgs-url string              BGS URL (default "https://bsky.network")
      --delete-all-posts            Whether to delete all posts from the index on startup (required for compliance with the EU right to be forgotten/GDPR article 17; deletions during uptime are handled using delete commits) (default true)
      --feed-generator-did string   DID of the feed generator (typically the hostname of the publicly reachable URL) (default "did:web:manager.atmosfeed.p8.lu")
      --feed-generator-url string   Publicly reachable URL of the feed generator (default "https://manager.atmosfeed.p8.lu")
  -h, --help                        help for manager
      --laddr string                Listen address (default ":1337")
      --limit int                   Maximum amount of posts to return for a feed (default 100)
      --origin string               Allowed CORS origin (default "https://atmosfeed.p8.lu")
      --ttl duration                Maximum age of posts to return for a feed (default 6h0m0s)

Global Flags:
      --postgres-url string   PostgreSQL URL (default "postgresql://postgres@localhost:5432/atmosfeed?sslmode=disable")
      --redis-url string      Redis URL (default "redis://localhost:6379/0")
      --s3-url string         S3 URL (default "http://minioadmin:minioadmin@localhost:9000?bucket=atmosfeed")
      --verbose               Whether to enable verbose logging

Worker

$ atmosfeed-server worker --help
Start an Atmosfeed worker

Usage:
  atmosfeed-server worker [flags]

Aliases:
  worker, w

Flags:
      --classifier-timeout duration   Amount of time after which to stop a classifier Scale function from running (default 1s)
  -h, --help                          help for worker
      --working-directory string      Working directory to use (default "/home/pojntfx/.local/share/atmosfeed/var/lib/atmosfeed")

Global Flags:
      --postgres-url string   PostgreSQL URL (default "postgresql://postgres@localhost:5432/atmosfeed?sslmode=disable")
      --redis-url string      Redis URL (default "redis://localhost:6379/0")
      --s3-url string         S3 URL (default "http://minioadmin:minioadmin@localhost:9000?bucket=atmosfeed")
      --verbose               Whether to enable verbose logging

Client CLI

$ atmosfeed-client --help
Create custom Bluesky feeds with WebAssembly and Scale.
Find more information at:
https://github.com/pojntfx/atmosfeed

Usage:
  atmosfeed-client [command]

Available Commands:
  apply           Create or update a feed on an Atmosfeed server
  completion      Generate the autocompletion script for the specified shell
  delete          Delete a feed from an Atmosfeed server
  delete-userdata Delete all user data from an Atmosfeed server
  dev             Develop a feed classifier locally
  export-userdata Export all user data from an Atmosfeed server
  help            Help about any command
  list            List published feeds on an Atmosfeed server
  publish         Publish a feed to a Bluesky PDS
  resolve         Resolve a handle to a DID
  unpublish       Unpublish a feed from a Bluesky PDS

Flags:
      --atmosfeed-url string   Atmosfeed server URL (default "https://manager.atmosfeed.p8.lu")
  -h, --help                   help for atmosfeed-client
      --password string        Bluesky password, preferably an app password (get one from https://bsky.app/settings/app-passwords)
      --pds-url string         PDS URL (default "https://bsky.social")
      --username string        Bluesky username (default "example.bsky.social")

Use "atmosfeed-client [command] --help" for more information about a command.

Expand subcommand reference

Apply

$ atmosfeed-client apply --help
Create or update a feed on an Atmosfeed server

Usage:
  atmosfeed-client apply [flags]

Aliases:
  apply, a

Flags:
      --clear-pinned              Whether to clear the pinned post field
      --feed-classifier string    Path to the feed classifier to upload (default "local-trending-latest.scale")
      --feed-rkey string          Machine-readable key for the feed (default "trending")
  -h, --help                      help for apply
      --pinned-feed-did string    DID of the pinned post for the feed (if left empty, no post will be pinned; empty values don't overwrite non-empty values, see --clear-pinned)
      --pinned-feed-rkey string   Machine-readable key of the pinned post for the feed (if left empty, no post will be pinned; empty values don't overwrite non-empty values, see --clear-pinned)

Global Flags:
      --atmosfeed-url string   Atmosfeed server URL (default "https://manager.atmosfeed.p8.lu")
      --password string        Bluesky password, preferably an app password (get one from https://bsky.app/settings/app-passwords)
      --pds-url string         PDS URL (default "https://bsky.social")
      --username string        Bluesky username (default "example.bsky.social")

Publish

$ atmosfeed-client publish --help
Publish a feed to a Bluesky PDS

Usage:
  atmosfeed-client publish [flags]

Aliases:
  publish, p

Flags:
      --feed-description string     Description for the feed (default "An example trending feed for Atmosfeed")
      --feed-generator-did string   DID of the feed generator (typically the hostname of the publicly reachable URL) (default "did:web:manager.atmosfeed.p8.lu")
      --feed-name string            Human-readable name for the feed) (default "Atmosfeed Trending")
      --feed-rkey string            Machine-readable key for the feed (default "trending")
  -h, --help                        help for publish

Global Flags:
      --atmosfeed-url string   Atmosfeed server URL (default "https://manager.atmosfeed.p8.lu")
      --password string        Bluesky password, preferably an app password (get one from https://bsky.app/settings/app-passwords)
      --pds-url string         PDS URL (default "https://bsky.social")
      --username string        Bluesky username (default "example.bsky.social")

List

$ atmosfeed-client list --help
List published feeds on an Atmosfeed server

Usage:
  atmosfeed-client list [flags]

Aliases:
  list, l

Flags:
  -h, --help   help for list

Global Flags:
      --atmosfeed-url string   Atmosfeed server URL (default "https://manager.atmosfeed.p8.lu")
      --password string        Bluesky password, preferably an app password (get one from https://bsky.app/settings/app-passwords)
      --pds-url string         PDS URL (default "https://bsky.social")
      --username string        Bluesky username (default "example.bsky.social")

Unpublish

$ atmosfeed-client unpublish --help
Unpublish a feed from a Bluesky PDS

Usage:
  atmosfeed-client unpublish [flags]

Aliases:
  unpublish, u

Flags:
      --feed-rkey string   Machine-readable key for the feed (default "trending")
  -h, --help               help for unpublish

Global Flags:
      --atmosfeed-url string   Atmosfeed server URL (default "https://manager.atmosfeed.p8.lu")
      --password string        Bluesky password, preferably an app password (get one from https://bsky.app/settings/app-passwords)
      --pds-url string         PDS URL (default "https://bsky.social")
      --username string        Bluesky username (default "example.bsky.social")

Delete

$ atmosfeed-client delete --help
Delete a feed from an Atmosfeed server

Usage:
  atmosfeed-client delete [flags]

Aliases:
  delete, d

Flags:
      --feed-rkey string   Machine-readable key for the feed (default "trending")
  -h, --help               help for delete

Global Flags:
      --atmosfeed-url string   Atmosfeed server URL (default "https://manager.atmosfeed.p8.lu")
      --password string        Bluesky password, preferably an app password (get one from https://bsky.app/settings/app-passwords)
      --pds-url string         PDS URL (default "https://bsky.social")
      --username string        Bluesky username (default "example.bsky.social")

Dev

$ atmosfeed-client dev --help
Develop a feed classifier locally

Usage:
  atmosfeed-client dev [flags]

Aliases:
  dev, v

Flags:
      --bgs-url string           BGS URL (default "https://bsky.network")
      --feed-classifier string   Path to the feed classifier to test (default "local-trending-latest.scale")
      --frontend-url string      Bluesky frontend URL to use when logging posts (default "https://bsky.app")
  -h, --help                     help for dev
      --max-posts int            Maximum amount of posts to store in memory before clearing the cache (default 1048576)
      --min-weight int           Minimum weight value the classifier has to return for a post to log it
      --quiet                    Whether to silently ignore any non-fatal errors (default true)
      --verbose                  Whether to enable verbose logging

Global Flags:
      --atmosfeed-url string   Atmosfeed server URL (default "https://manager.atmosfeed.p8.lu")
      --password string        Bluesky password, preferably an app password (get one from https://bsky.app/settings/app-passwords)
      --pds-url string         PDS URL (default "https://bsky.social")
      --username string        Bluesky username (default "example.bsky.social")

Resolve

$ atmosfeed-client resolve --help
Resolve a handle to a DID

Usage:
  atmosfeed-client resolve [flags]

Aliases:
  resolve, p

Flags:
      --handle string   Handle/username/domain to resolve
  -h, --help            help for resolve

Global Flags:
      --atmosfeed-url string   Atmosfeed server URL (default "https://manager.atmosfeed.p8.lu")
      --password string        Bluesky password, preferably an app password (get one from https://bsky.app/settings/app-passwords)
      --pds-url string         PDS URL (default "https://bsky.social")
      --username string        Bluesky username (default "example.bsky.social")

Export User Data

$ atmosfeed-client export-userdata --help
Export all user data from an Atmosfeed server

Usage:
  atmosfeed-client export-userdata [flags]

Aliases:
  export-userdata, eu

Flags:
  -h, --help         help for export-userdata
      --out string   Directory to export user data to (default "atmosfeed-userdata")

Global Flags:
      --atmosfeed-url string   Atmosfeed server URL (default "https://manager.atmosfeed.p8.lu")
      --password string        Bluesky password, preferably an app password (get one from https://bsky.app/settings/app-passwords)
      --pds-url string         PDS URL (default "https://bsky.social")
      --username string        Bluesky username (default "example.bsky.social")

Delete User Data

$ atmosfeed-client delete-userdata --help
Delete all user data from an Atmosfeed server

Usage:
  atmosfeed-client delete-userdata [flags]

Aliases:
  delete-userdata, du

Flags:
  -h, --help   help for delete-userdata

Global Flags:
      --atmosfeed-url string   Atmosfeed server URL (default "https://manager.atmosfeed.p8.lu")
      --password string        Bluesky password, preferably an app password (get one from https://bsky.app/settings/app-passwords)
      --pds-url string         PDS URL (default "https://bsky.social")
      --username string        Bluesky username (default "example.bsky.social")

Environment Variables

All command line arguments described above can also be set using environment variables; for example, to set --pds-url to https://bsky.social with an environment variable, use ATMOSFEED_PDS_URL='https://bsky.social'. In addition to this, there are also some aliases in place for compatibility with Railway/Heroku conventions; for example, you can export PORT=3000 to listen on port 3000 as an alternative to setting the listen address, or export DATABASE_URL='postgresql://username:password@myhost:5432/mydb' to set the database URL. For more info, see the command line arguments.

Acknowledgements

loopholelabs/scale provides the WebAssembly-based plugin system.
sqlc-dev/sqlc provides the SQL library.
pressly/goose provides migration support.
bluesky-social/indigo provides the Bluesky API client.

Contributing

To contribute, please use the GitHub flow and follow our Code of Conduct.

To build and start a development version of Atmosfeed locally, run the following:

# Download the source code
git clone https://github.com/pojntfx/atmosfeed.git
cd atmosfeed

# Setup dependencies
docker rm -f atmosfeed-postgres && docker run -d --name atmosfeed-postgres -p 5432:5432 -e POSTGRES_HOST_AUTH_METHOD=trust -e POSTGRES_DB=atmosfeed postgres
docker rm -f atmosfeed-redis && docker run --name atmosfeed-redis -p 6379:6379 -d redis
docker rm -f atmosfeed-minio && docker run --name atmosfeed-minio -p 9000:9000 -d minio/minio server /data

make -j$(nproc) depend

# Start manager
export ATMOSFEED_ORIGIN='http://localhost:3000'
export ATMOSFEED_FEED_GENERATOR_DID='did:web:atmosfeed.serveo.net'
make -j$(nproc) depend/cli && go run ./cmd/atmosfeed-server manager

# Start a tunnel to reach the manager from the public internet
ssh -R atmosfeed.serveo.net:80:localhost:1337 serveo.net

# Start worker(s)
make -j$(nproc) depend/cli && go run ./cmd/atmosfeed-server worker --working-directory ~/.local/share/atmosfeed/var/lib/atmosfeed/worker-1
make -j$(nproc) depend/cli && go run ./cmd/atmosfeed-server worker --working-directory ~/.local/share/atmosfeed/var/lib/atmosfeed/worker-2

# Download example feeds
curl -Lo out/local-everything-latest.scale https://github.com/pojntfx/bluesky-feeds/releases/download/release-main/local-everything-latest.scale
curl -Lo out/local-german-latest.scale https://github.com/pojntfx/bluesky-feeds/releases/download/release-main/local-german-latest.scale
curl -Lo out/local-question-latest.scale https://github.com/pojntfx/bluesky-feeds/releases/download/release-main/local-question-latest.scale
curl -Lo out/local-trending-latest.scale https://github.com/pojntfx/bluesky-feeds/releases/download/release-main/local-trending-latest.scale

# Deploy example feeds
export ATMOSFEED_PASSWORD='asdf'
export ATMOSFEED_USERNAME='pojntfxtesting.bsky.social'
export ATMOSFEED_ATMOSFEED_URL='http://localhost:1337'
export ATMOSFEED_FEED_GENERATOR_DID='did:web:atmosfeed.serveo.net'

go run ./cmd/atmosfeed-client/ apply --feed-rkey everything --feed-classifier ./out/local-everything-latest.scale
go run ./cmd/atmosfeed-client/ publish --feed-rkey everything --feed-name 'Atmosfeed Everything' --feed-description 'Newest posts on Bluesky (testing feed)'

go run ./cmd/atmosfeed-client/ apply --feed-rkey questions --feed-classifier ./out/local-questions-latest.scale
go run ./cmd/atmosfeed-client/ publish --feed-rkey questions --feed-name 'Atmosfeed Questions' --feed-description 'Most popular questions on Bluesky in the last 24h (testing feed).'

go run ./cmd/atmosfeed-client/ apply --feed-rkey german --feed-classifier ./out/local-german-latest.scale
go run ./cmd/atmosfeed-client/ publish --feed-rkey german --feed-name 'Atmosfeed German' --feed-description 'Most popular German posts on Bluesky in the last 24h (testing feed)'

go run ./cmd/atmosfeed-client/ apply --feed-rkey trending --feed-classifier ./out/local-trending-latest.scale
go run ./cmd/atmosfeed-client/ publish --feed-rkey trending --feed-name 'Atmosfeed Trending' --feed-description 'Most popular trending posts on Bluesky in the last 24h (testing feed)'

# Remove example feeds
go run ./cmd/atmosfeed-client/ unpublish --feed-rkey questions
go run ./cmd/atmosfeed-client/ delete --feed-rkey questions

go run ./cmd/atmosfeed-client/ unpublish --feed-rkey german
go run ./cmd/atmosfeed-client/ delete --feed-rkey german

go run ./cmd/atmosfeed-client/ unpublish --feed-rkey everything
go run ./cmd/atmosfeed-client/ delete --feed-rkey everything

go run ./cmd/atmosfeed-client/ unpublish --feed-rkey trending
go run ./cmd/atmosfeed-client/ delete --feed-rkey trending

# Start frontend
cd frontend
bun dev # Now visit http://localhost:3000 to open the frontend and sign in

# Export or delete user data for privacy & interoperability
go run ./cmd/atmosfeed-client/ export-userdata --out ./out/atmosfeed-userdata
go run ./cmd/atmosfeed-client/ delete-userdata

Have any questions or need help? Chat with us on Matrix!

License

SPDX-License-Identifier: Apache-2.0