textile/buckets repository overview

⁠DEPRECATION NOTICE: Textile's hosted Hub infrastructure will be taken off-line on January 9th, 2023. At this time, all ThreadDB and Bucket data will no longer be available, and will subsequently be removed. See https://github.com/textileio/textile/issues/578⁠ for further details.

⁠textile

Textile connects and extends Libp2p⁠, IPFS⁠, and Filecoin⁠. Three interoperable technologies makeup Textile:

• ThreadDB⁠: A server-less p2p database built on Libp2p
• Powergate⁠: File storage built on Filecoin and IPFS
• Buckets⁠: File and dynamic directory storage built on ThreadDB, Powergate, and UnixFS⁠.

Join us on our public Slack channel⁠ for news, discussions, and status updates. Check out our blog⁠ for the latest posts and announcements.

⁠Table of Contents

• textile⁠
  • Security⁠
  • Background⁠
  • Install⁠
    • The Hub⁠
      • hubd⁠
      • hub⁠
    • Buckets⁠
      • buckd⁠
      • buck⁠
    • The Buckets Library⁠
  • Getting Started⁠
    • The Hub⁠
    • Running Buckets⁠
    • Creating a bucket⁠
    • Creating a private bucket⁠
    • Adding files and folders to a bucket⁠
    • Recreating an existing bucket⁠
    • Creating a bucket from an existing Cid⁠
    • Exploring bucket contents⁠
    • Resetting bucket contents⁠
    • Watching a bucket for changes⁠
    • Protecting a file with a password⁠
    • Sharing bucket files and folders⁠
    • Creating a Filecoin bucket archive⁠
    • Multi-writer buckets⁠
    • Deleting a bucket⁠
    • Using the Buckets Library⁠
      • Creating a bucket⁠
      • Getting an existing bucket⁠
      • Pushing local files⁠
      • Pulling remote changes⁠
    • Using the Mail Library⁠
      • Creating a mailbox⁠
      • Getting an existing mailbox⁠
      • Sending a message⁠
      • Watching for new messages⁠
  • Developing⁠
  • Contributing⁠
  • Changelog⁠
  • License⁠

⁠Security

Textile is still under heavy development and no part of it should be used before a thorough review of the underlying code and an understanding APIs and protocols may change rapidly. There may be coding mistakes, and the underlying protocols may contain design flaws. Please let us know⁠ immediately if you have discovered a security vulnerability.

Please also read the security note⁠ for go-ipfs⁠.

⁠Background

Go to the docs⁠ for more about the motivations behind Textile.

⁠Install

This repo contains two service daemons with CLIs and a Buckets Library for building local-first apps and services.

⁠The Hub

⁠hubd

• Prebuilt package: See release assets⁠
• Docker image: See the latest tag on Docker Hub⁠
• Build from the source:

git clone https://github.com/textileio/textile
cd textile
go get ./cmd/hubd

⁠hub

• Prebuilt package: See release assets⁠
• Build from the source:

git clone https://github.com/textileio/textile
cd textile
go get ./cmd/hub

Note: hub includes buck as a subcommand: hub buck. This is because hubd hosts buckd, along with other services.

hub is built in part on the gRPC client⁠, which can be imported to an existing project:

import "github.com/textileio/textile/v2/api/hub/client"

⁠Buckets

⁠buckd

• Prebuilt package: See release assets⁠
• Docker image: See the buckets tag on Docker Hub⁠
• Build from the source:

git clone https://github.com/textileio/textile
cd textile
go get ./cmd/buckd

⁠buck

• Prebuilt package: See release assets⁠
• Build from the source:

git clone https://github.com/textileio/textile
cd textile
go get ./cmd/buck

buck is built in part on the gRPC client⁠, which can be imported in an existing project:

import "github.com/textileio/textile/v2/api/buckets/client"

⁠The Buckets Library

import "github.com/textileio/textile/v2/buckets/local"

The full spec is available here⁠.

⁠Getting Started

⁠The Hub

The Hub daemon (hubd), a.k.a. The Hub, is a hosted wrapper around other Textile services that includes developer accounts for individuals and organizations. You are encouraged to run your own, and we strongly discourage the use of the hosted Textile Hub⁠ as it will soon be shutting down.

The layout of the hub client CLI mirrors the services wrapped by hubd:

• hub threads provides limited access to ThreadDB.
• hub buck provides access to Buckets (buckd) by wrapping the standalone buck CLI.
• hub buck archive provides limited access to The Hub's hosted Powergate instance, and the Filecoin network.

Try hub --help for more usage.

The Hub Client.

Usage:
  hub [command]

Available Commands:
  billing     Billing management
  buck        Manage an object storage bucket
  destroy     Destroy your account
  fil         Interact with Filecoin related commands.
  help        Help about any command
  init        Initialize account
  keys        API key management
  login       Login
  logout      Logout
  orgs        Org management
  threads     Thread management
  update      Update the hub CLI
  version     Show current version
  whoami      Show current user

Flags:
      --api string        API target (default "api.hub.textile.io:443")
  -h, --help              help for hub
      --identity string   User identity
      --key string        User API key
      --newIdentity       Generate a new user identity
  -o, --org string        Org username
      --secret string     User API secret
  -s, --session string    User session token
      --token string      User identity token

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

Read more about The Hub, including how to create an account⁠, in the docs⁠.

⁠Running Buckets

Much like threadsd⁠, the buckd daemon can be run as a server or alongside desktop apps or command-line tools. The easiest way to run buckd is by using the provided Docker Compose files. If you're new to Docker and/or Docker Compose, get started here⁠. Once you are setup, you should have docker-compose in your PATH.

Create an .env file and add the following values:

REPO_PATH=~/myrepo
BUCK_LOG_DEBUG=true

Copy this compose file⁠ and run it with the following command.

docker-compose -f docker-compose.yml up

Congrats! Now you have Buckets running locally.

The Docker Compose file starts an IPFS node, which is used to pin bucket files and folders. You could point buckd to a different (possibly remote) IPFS node by setting the BUCK_ADDR_IPFS_API variable to a different multiaddress.

By default, this approach does not start Powergate⁠. If you do, be sure to set the BUCK_ADDR_POWERGATE_API variable to the multiaddress of your Powergate. buckd must be configured with Powergate to enable Filecoin archiving with buck archive.

⁠Creating a bucket

Since hub buck and buck are functionally identical, this section will focus on buck and the Buckets Library using a locally running buckd.

First off, take a look at buck --help.

The Bucket Client.

Manages files and folders in an object storage bucket.

Usage:
  buck [command]

Available Commands:
  add         Add a UnixFs DAG locally at path
  archive     Create a Filecoin archive
  cat         Cat bucket objects at path
  decrypt     Decrypt bucket objects at path with password
  destroy     Destroy bucket and all objects
  encrypt     Encrypt file with a password
  help        Help about any command
  init        Initialize a new or existing bucket
  links       Show links to where this bucket can be accessed
  ls          List top-level or nested bucket objects
  pull        Pull bucket object changes
  push        Push bucket object changes
  root        Show bucket root CIDs
  status      Show bucket object changes
  watch       Watch auto-pushes local changes to the remote

Flags:
      --api string   API target (default "127.0.0.1:3006")
  -h, --help         help for buck

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

A Textile bucket functions a bit like an S3 bucket. It's a virtual filesystem where you can push, pull, list, and cat files. You can share them via web links or render the whole thing as a website or web app. They also function a bit like a Git repository. The point of entry is from a folder on your local machine that is synced to a remote.

To get started, initialize a new bucket.

mkdir mybucket && cd mybucket
buck init

When prompted, give your bucket a name and either opt-in or decline bucket encyption (see Creating a private bucket⁠ for more about bucket encryption).

You should now see two links for the new bucket on the locally running gateway.

> http://127.0.0.1:8006/thread/bafkq3ocmdkrljadlgybtvocytpdw4hbnzygxecxehdp7pfj32lxp34a/buckets/bafzbeifyzfm3kosie25s5qthvvcjrr42ivd7doqhwvu5m4ks7uqv4j5lyi Thread link
> http://127.0.0.1:8006/ipns/bafzbeifyzfm3kosie25s5qthvvcjrr42ivd7doqhwvu5m4ks7uqv4j5lyi IPNS link (propagation can be slow)
> Success! Initialized /path/to/mybucket as a new empty bucket

The first URL is the link to the ThreadDB instance. Internally, a collection named buckets is created. Each new instance in this collection amounts to a new bucket. However, when you visit this link, you'll notice a custom file browser. This is because the gateway considers the built-in buckets collection a special case. You can still view the raw ThreadDB instance by appending ?json=true to the URL.

The second URL is the bucket's unique IPNS address, which is auto-updated when you add, modify, or delete files.

If you have configured the daemon with DNS settings, you will see a third URL that links to the bucket's WWW address, where it is rendered as a static website / client-side application. See buckd --help for more info.

Important: If your bucket is private (encrypted), an access token (JWT) will be appended to these links. This token represents your identity across all buckets and should not be shared without caution.

buck init created a configuration folder in mybucket called .textile. This folder is somewhat like a .git folder, as it contains information about the bucket's remote address and local state.

.textile/config.yml will look something like,

key: bafzbeifyzfm3kosie25s5qthvvcjrr42ivd7doqhwvu5m4ks7uqv4j5lyi
thread: bafkq3ocmdkrljadlgybtvocytpdw4hbnzygxecxehdp7pfj32lxp34a

Where key is the bucket's unique key, and thread is it's ThreadDB ID.

Additionally, .textile/repo contains a repository describing the current file structure, which is used to stage changes against the remote.

⁠Creating a private bucket

Bucket encryption (AES-CTR + AES-512 HMAC) happens entirely within the buckd, meaning your data gets encrypted on the way in, and decrypted on the way out. This type of encryption has two goals:

• Obfuscate bucket data / files (the normal goal of encryption)
• Obfuscate directory structure, which amounts to encrypting IPLD⁠ nodes and their links.

As a result of these goals, we refer to encrypted buckets as private buckets. Read more about bucket encryption here⁠.

To create a new private bucket, use the --private flag with buck init or respond y when prompted.

In addition to bucket-level encryption, you can also protect a file with a password⁠.

⁠Adding files and folders to a bucket

Bucket files and folders are content-addressed by Cids. Check out the spec⁠ if you're unfamiliar with Cids.

buck stages new files as additions:

echo "hello world" > hello.txt
buck status
> new file:  hello.txt

buck status is powered by DAG-based diffing. Much like git, this allows buck to only push and pull changes. Read more about bucket diffing in the docs⁠, or check out this in-depth blog post⁠.

Use push to sync the change.

buck push
+ hello.txt: bafkreifjjcie6lypi6ny7amxnfftagclbuxndqonfipmb64f2km2devei4
> bafybeihm4zrnrsdroazwsvk3i65ooqzdftaugdkjiedr6ocq65u3ap4wni

The output shows the Cid of the added file and the bucket's new root Cid.

push will sync all types of file changes: Additions, modifications, and deletions.

⁠Recreating an existing bucket

It's often useful to recreate a bucket from the remote. This is somewhat like re-cloning a Git repo. This can be done in a different location on the same machine, or, if buckd has a public IP address, from a completely different machine.

Let's recreate the bucket from the previous step in a new directory outside of the original bucket.

mkdir mybucket2 && cd mybucket2
buck init --existing

The --existing flag allows for interactively selecting an existing bucket to initialize from.

? Which exiting bucket do you want to init from?:
  ▸ MyBucket bafzbeifyzfm3kosie25s5qthvvcjrr42ivd7doqhwvu5m4ks7uqv4j5lyi

At this point, there's only one bucket to choose from.

Note: If buckd was running inside The Hub (hubd), you would be able to choose from buckets belonging to your Organizations and well as your individual Developer account by using the --org flag. Read more about Hub Accounts and Organizations here⁠.

> Selected bucket MyBucket
+ hello.txt: bafkreifjjcie6lypi6ny7amxnfftagclbuxndqonfipmb64f2km2devei4
+ .textileseed: bafkreifbdzttoqsch5j66hfmcbsic6qvwrikibgzfbg3tn7rc3j63ukk3u
> Your bucket links:
> http://127.0.0.1:8006/thread/bafkq3ocmdkrljadlgybtvocytpdw4hbnzygxecxehdp7pfj32lxp34a/buckets/bafzbeifyzfm3kosie25s5qthvvcjrr42ivd7doqhwvu5m4ks7uqv4j5lyi Thread link
> http://127.0.0.1:8006/ipns/bafzbeifyzfm3kosie25s5qthvvcjrr42ivd7doqhwvu5m4ks7uqv4j5lyi IPNS link (propagation can be slow)
> Success! Initialized /path/to/mybucket2 from an existing bucket

Just as before, the output shows the bucket's remote links. However, in this case init also pulled down the content.

Note: .textileseed is used to randomize a bucket's top level Cid and cannot be modified.

The --existing flag is really just a helper that sets the --thread and --key flags, which match the config values we saw earlier. We could have used those flags directly to achieve the same result.

buck init --thread bafkq3ocmdkrljadlgybtvocytpdw4hbnzygxecxehdp7pfj32lxp34a --key bafzbeifyzfm3kosie25s5qthvvcjrr42ivd7doqhwvu5m4ks7uqv4j5lyi

Lastly, we could have just copied .textile/config.yml to a new directory and used buck pull to pull down the existing content.

⁠Creating a bucket from an existing Cid

Sometimes it's useful to create a bucket from a UnixFS⁠ directory that is already on the IPFS network.

We can simulate this scenario by adding a local folder to IPFS and then using its root Cid to create a bucket with the --cid flag. Here's a local directory.

.
├── a
│   ├── bar.txt
│   ├── foo.txt
│   └── one
│       ├── baz.txt
│       ├── buz.txt
│       └── two
│           ├── boo.txt
│           └── fuz.txt
├── b
│   ├── foo.txt
│   └── one
│       ├── baz.txt
│       ├── muz.txt
│       ├── three
│       │   └── far.txt
│       └── two
│           └── fuz.txt
└── c
    ├── one.jpg
    └── two.jpg

Use the recursvie flag -r with ipfs add.

ipfs add -r .
added QmcDkcMJXZsNnExehsE1Yh6SRWucHa9ruVT82gpL83431W mydir/a/bar.txt
added QmYiUq2U6euWnKag23wFppG12hon4EBDswdoe4MwrKzDBn mydir/a/foo.txt
added QmXrd35ja3kknnmgj5kyDM74jfG8GLJJQGtRpEQpXCLTR3 mydir/a/one/baz.txt
added QmSWJvCzotB3CbdxVu8mBvmLqpSuEQgUoJHTFy1azRfwhT mydir/a/one/buz.txt
added QmT6h1eaBV74Sh75upE7ugFLkBnmyGr3WsQ8w8yx5NjgPV mydir/a/one/two/boo.txt
added QmTdg1b5eWEx4zJtrgvew1inkkZ29fp9mbQ4uHyKurW8Ub mydir/a/one/two/fuz.txt
added QmYiQAk1seXrmuQkpGE83AxJyNZDK1RNSaLyp3Z4r1zsrB mydir/b/foo.txt
added QmXrd35ja3kknnmgj5kyDM74jfG8GLJJQGtRpEQpXCLTR3 mydir/b/one/baz.txt
added QmSWJvCzotB3CbdxVu8mBvmLqpSuEQgUoJHTFy1azRfwhT mydir/b/one/muz.txt
added QmYs12A3CGSTHX4QrsvBe2AvLHEThrapXoTFQpyh8AzpFa mydir/b/one/three/far.txt
added QmTdg1b5eWEx4zJtrgvew1inkkZ29fp9mbQ4uHyKurW8Ub mydir/b/one/two/fuz.txt
added QmaLpwNPwftSQY3w4ZtMfZ8k38D5EgK2bcDuU4UwzREJpi mydir/c/one.jpg
added QmYLiWv2WXQd1m8YyHx4dMoj8B3Kuiuu7pCCoYibkqKyVj mydir/c/two.jpg
added QmT5YXeCfbMuVjanbHjQhECUQSACJLecfmjRBZHvmu5FDU mydir/a/one/two
added QmWh2Wx9Lec4wbEvFbsq4HmYjFmgUFtxNJ8wEVwXjhJ2uk mydir/a/one
added QmSujVHvG8Y3Jv21AbMFNQPphjyqNamh6cvdyXSD1jAtSZ mydir/a
added QmUGSorWDy2JiKYvQuJzEb4TnYDuDNLcdFyR6NhMwnwdvy mydir/b/one/three
added QmWvX7UVexbjXJtxKMyMSgGpPesFQD7teNTqUcDsP2mzW6 mydir/b/one/two
added QmPyMD67EgSZS1WpvgudHkxbA5zgjqmse8srPpFb9sVefT mydir/b/one
added QmQdAtg5NkwkvLtTbka3eci58UGj3m9AehC2sbksGSbjPZ mydir/b
added QmcjtVAF9PQfMKTc57vcvZeBrzww3TLxPcQfUQW7cXXLJL mydir/c
added QmcvkGF2t8Z94UqhdtdFRokGoqypbGyKkzRPVF4owmjVrE mydir

After adding the entire directory, we see the root Cid is QmcvkGF2t8Z94UqhdtdFRokGoqypbGyKkzRPVF4owmjVrE. Let's create the bucket using this Cid.

buck init --cid QmcvkGF2t8Z94UqhdtdFRokGoqypbGyKkzRPVF4owmjVrE

The files behind the Cid will be pulled into the new bucket.

+ a/bar.txt: QmcDkcMJXZsNnExehsE1Yh6SRWucHa9ruVT82gpL83431W
+ a/foo.txt: QmYiUq2U6euWnKag23wFppG12hon4EBDswdoe4MwrKzDBn
+ a/one/two/fuz.txt: QmTdg1b5eWEx4zJtrgvew1inkkZ29fp9mbQ4uHyKurW8Ub
+ a/one/baz.txt: QmXrd35ja3kknnmgj5kyDM74jfG8GLJJQGtRpEQpXCLTR3
+ c/two.jpg: QmYLiWv2WXQd1m8YyHx4dMoj8B3Kuiuu7pCCoYibkqKyVj
+ b/foo.txt: QmYiQAk1seXrmuQkpGE83AxJyNZDK1RNSaLyp3Z4r1zsrB
+ a/one/buz.txt: QmSWJvCzotB3CbdxVu8mBvmLqpSuEQgUoJHTFy1azRfwhT
+ a/one/two/boo.txt: QmT6h1eaBV74Sh75upE7ugFLkBnmyGr3WsQ8w8yx5NjgPV
+ b/one/muz.txt: QmSWJvCzotB3CbdxVu8mBvmLqpSuEQgUoJHTFy1azRfwhT
+ b/one/three/far.txt: QmYs12A3CGSTHX4QrsvBe2AvLHEThrapXoTFQpyh8AzpFa
+ b/one/baz.txt: QmXrd35ja3kknnmgj5kyDM74jfG8GLJJQGtRpEQpXCLTR3
+ b/one/two/fuz.txt: QmTdg1b5eWEx4zJtrgvew1inkkZ29fp9mbQ4uHyKurW8Ub
+ c/one.jpg: QmaLpwNPwftSQY3w4ZtMfZ8k38D5EgK2bcDuU4UwzREJpi
> Your bucket links:
> http://127.0.0.1:8006/thread/bafk3k3itq2rsybcvhf6wuvumruw3j6cw7ixhrtx4ek45qgvp3e7u2xa/buckets/bafzbeiawo6ghgsqjlorii4wghdl4tzz54x2kiwtcgtaq7b3h5gta2yok2i Thread link
> http://127.0.0.1:8006/ipns/bafzbeiawo6ghgsqjlorii4wghdl4tzz54x2kiwtcgtaq7b3h5gta2yok2i IPNS link (propagation can be slow)
> Success! Initialized /path/to/mybucket3 as a new bootstrapped bucket

Currently, UnixFS in go-ipfs uses Cid version 0, which is why we see all these old-style Cids started with Qm. Of course, you can also use UnixFS directories that use Cid version 1.

Similar to initializing a new bucket from an existing Cid, buck add allows you to add and/or merge in an existing UnixFS directory to an existing bucket. Like adding new files locally, this works by pulling down the UnixFS content from the IPFS network into the local bucket. Sync the changes with buck push as normal.

Pulling an existing UnixFS directory into a new or existing private bucket is also possible. Just opt-in to encryption during initialization as normal. buckd will recursively encrypt (without duplicating) the Cid's IPLD file and directory nodes as they are pulled into the new bucket.

⁠Exploring bucket contents

Use buck ls [path] to explore bucket contents. Omitting [path] will list the top-level directory.

buck ls

  NAME          SIZE     DIR    OBJECTS  CID
  .textileseed  32       false  n/a      bafkreiezexkrnk7yew6glm6sulhur66bbecc2aeaitf7uz4ymmp442lepu
  a             3726     true   3        QmSujVHvG8Y3Jv21AbMFNQPphjyqNamh6cvdyXSD1jAtSZ
  b             3191     true   2        QmQdAtg5NkwkvLtTbka3eci58UGj3m9AehC2sbksGSbjPZ
  c             1537626  true   2        QmcjtVAF9PQfMKTc57vcvZeBrzww3TLxPcQfUQW7cXXLJL

Use [path] to drill into directories, e.g.,

buck ls a

  NAME     SIZE  DIR    OBJECTS  CID
  bar.txt  517   false  n/a      QmcDkcMJXZsNnExehsE1Yh6SRWucHa9ruVT82gpL83431W
  foo.txt  557   false  n/a      QmYiUq2U6euWnKag23wFppG12hon4EBDswdoe4MwrKzDBn
  one      2502  true   3        QmWh2Wx9Lec4wbEvFbsq4HmYjFmgUFtxNJ8wEVwXjhJ2uk

buck cat functions a lot like ls, but cats file contents to stdout.

⁠Resetting bucket contents

Similar to a git reset --hard, you can use buck pull --hard to discard local changes that have not been pushed.

Continuing with the bucket above, add, modify, and/or delete some files. buck status should show your staged changes.

buck status
> modified:  a/bar.txt
> deleted:   a/one/baz.txt
> new file:  b/one/three/car.txt
> deleted:   b/foo.txt

Normally, buck pull will move your local changes to temporary .buckpatch files, apply the remote / upstream changes, then reapply your local changes. However, the --hard flag will prune all local changes, resetting the local bucket contents to match the remote exactly.

buck pull --hard
+ a/one/baz.txt: QmXrd35ja3kknnmgj5kyDM74jfG8GLJJQGtRpEQpXCLTR3
+ b/foo.txt: QmYiQAk1seXrmuQkpGE83AxJyNZDK1RNSaLyp3Z4r1zsrB
+ a/bar.txt: QmcDkcMJXZsNnExehsE1Yh6SRWucHa9ruVT82gpL83431W
- b/one/three/car.txt
> QmTz6HoC18QQqAEtYhfLc4Fse3LPbSCKV8vouvE88MKjFj

Now buck status will report > Everything up-to-date.

Try buck pull --help for more options when pulling the remote.

⁠Watching a bucket for changes

So far we've seen how a bucket can change locally, but the remote can also change. This could happen for a couple reasons:

• Changes are pushed from a different bucket copy against the same buckd.
• Changes are pushed from a different buckd at the ThreadDB layer. This is known as a multi-writer scenario. See Multi-writer buckets⁠ for more.

In either case, it is possible to listen for and apply the remote changes using buck watch. This will also watch for local changes and auto-push them to the remote. In this way, multiple copies of the same bucket can be kept in sync.

watch will block until it's