#sqlite #web-apps #embedded-database #command-line #database-server #web-api #sql-database

bin+lib ayb

ayb makes it easy to create, host, and share embedded databases like SQLite and DuckDB

8 releases

0.1.8 Aug 27, 2024
0.1.7 Feb 25, 2024
0.1.6 Jan 14, 2024
0.1.5 Nov 18, 2023
0.1.3 Apr 28, 2023

#1203 in Web programming

Download history 22/week @ 2024-07-27 155/week @ 2024-08-24 10/week @ 2024-08-31 28/week @ 2024-09-21 4/week @ 2024-09-28 1/week @ 2024-10-12

441 downloads per month

Apache-2.0

175KB
4K SLoC

ayb

ayb makes it easy to create databases, share them with collaborators, and query them from a web application or the command line.

With ayb, all your (data)base can finally belong to you. Move SQL for great justice.

Build status

Introduction

ayb is a database management system with easy-to-host instances that enable users to quickly register an account, create databases, share them with collaborators, and query them from a web application or the command line. An ayb server allows users to create SQLite databases (other databases to come), and then exposes those databases through an HTTP API.

To learn more about why ayb matters, how it works, or who it's for, read this introductory blog post.

alpha warning: ayb is neither feature complete nor production-ready. Functionality like authentication, permissions, collaboration, isolation, high availability, and transaction support are on the Roadmap but not available today. I work on ayb as a hobbyist side project.

Getting started

Installing

ayb is written in Rust, and is available as the ayb crate. Assuming you have installed Rust on your machine, installing ayb takes a single command:

cargo install ayb

Running a server

An ayb server stores its metadata in SQLite or PostgreSQL, and stores the databases it's hosting on a local disk. An ayb.toml file tells the server what host/port to listen for connections on, how to connect to the database, and the data path for the hosted databases. You can generate a starter file with ayb default_server_config.

$ ayb default_server_config > ayb.toml

$ cat ayb.toml

host = "0.0.0.0"
port = 5433
database_url = "sqlite://ayb_data/ayb.sqlite"
# Or, for Postgres:
# database_url = "postgresql://postgres_user:test@localhost:5432/test_db"
data_path = "./ayb_data"

[authentication]
# A secret (and unique to your server) key that is used for account registration.
fernet_key = "<UNIQUE_KEY_GENERATED_BY_COMMAND>="
token_expiration_seconds = 3600

[email]
from = "Server Sender <server@example.org>"
reply_to = "Server Reply <replyto@example.org>"
smtp_host = "localhost"
smtp_port = 465
smtp_username = "login@example.org"
smtp_password = "the_password"

Running the server then requires one command

$ ayb server

Running a client

Once the server is running, you can register a user (in this case, marcua), create a database marcua/test.sqlite, and issue SQL as you like. Here's how to do that at the command line:

$ ayb client --url http://127.0.0.1:5433 register marcua you@example.com
Check your email to finish registering marcua

# You will receive an email at you@example.com instructing you to type the next command
$ ayb client confirm <TOKEN_FROM_EMAIL>
Successfully authenticated and saved token <API_TOKEN>

$ ayb client create_database marcua/test.sqlite
Successfully created marcua/test.sqlite

$ ayb client list marcua
 Database slug | Type 
---------------+--------
 test.sqlite   | sqlite 

$ ayb client query marcua/test.sqlite "CREATE TABLE favorite_databases(name varchar, score integer);"

Rows: 0

# If you don't pass a query to the query command, ayb launches an interactive query session
$ ayb client query marcua/test.sqlite
Launching an interactive session for marcua/test.sqlite
marcua/test.sqlite> INSERT INTO favorite_databases (name, score) VALUES ("PostgreSQL", 10);

Rows: 0
marcua/test.sqlite> INSERT INTO favorite_databases (name, score) VALUES ("SQLite", 9);

Rows: 0
marcua/test.sqlite> INSERT INTO favorite_databases (name, score) VALUES ("DuckDB", 9);

Rows: 0
marcua/test.sqlite> SELECT * FROM favorite_databases;
 name       | score 
------------+-------
 PostgreSQL | 10 
 SQLite     | 9 
 DuckDB     | 9 

Rows: 3
marcua/test.sqlite>

$ ayb client update_profile marcua --display_name 'Adam Marcus' --links 'http://marcua.net'

Successfully updated profile

$ ayb client profile marcua
 Display name | Description | Organization | Location | Links 
--------------+-------------+--------------+----------+-------------------
 Adam Marcus  |             |              |          | http://marcua.net 

Note that the command line also saved a configuration file for your convenience so you don't have to keep entering a server URL or API token. If you ever want to set these explicitly, the --url/--token command-line flags and AYB_SERVER_URL/AYB_API_TOKEN environment variables will override whatever is in the saved configuration. By default, the configuration file can be found in:

  • Linux: /home/alice/.config/ayb/ayb.json
  • MacOS (untested): /Users/Alice/Library/Application Support/org.ayb.ayb/ayb.json
  • Windows (untested): C:\Users\Alice\AppData\Roaming\ayb\ayb\config\ayb.json

The command line invocations above are a thin wrapper around ayb's HTTP API. Here are the same commands as above, but with curl:

$ curl -w "\n" -X POST http://127.0.0.1:5433/v1/register -H "entity-type: user" -H "entity: marcua" -H "email-address: your@example.com"

{}

$ curl -w "\n" -X POST http://127.0.0.1:5433/v1/confirm -H "authentication-token: TOKEN_FROM_EMAIL"

{"entity":"marcua","token":"<API_TOKEN>"}

$ curl -w "\n" -X POST http://127.0.0.1:5433/v1/marcua/test.sqlite/create -H "db-type: sqlite" -H "authorization: Bearer <API_TOKEN_FROM_PREVIOUS_COMMAND>"

{"entity":"marcua","database":"test.sqlite","database_type":"sqlite"}

$ curl -w "\n" -X PATCH http://127.0.0.1:5433/v1/entity/marcua -H "authorization: Bearer <API_TOKEN_FROM_PREVIOUS_COMMAND>" -d "{\"display_name\": \"Adam Marcus\"}"

{}

$ curl -w "\n" -X GET http://localhost:5433/v1/entity/marcua -H "authorization: Bearer <API_TOKEN_FROM_PREVIOUS_COMMAND>"

{"slug":"marcua","databases":[{"slug":"test.sqlite","database_type":"sqlite"}],"profile":{"display_name":"Adam Marcus"}}

$ curl -w "\n" -X POST http://127.0.0.1:5433/v1/marcua/test.sqlite/query -H "authorization: Bearer <API_TOKEN_FROM_PREVIOUS_COMMAND>" -d 'CREATE TABLE favorite_databases(name varchar, score integer);'

{"fields":[],"rows":[]}

$ curl -w "\n" -X POST http://127.0.0.1:5433/v1/marcua/test.sqlite/query -H "authorization: Bearer <API_TOKEN_FROM_PREVIOUS_COMMAND>" -d "INSERT INTO favorite_databases (name, score) VALUES (\"PostgreSQL\", 10);"

{"fields":[],"rows":[]}

$ curl -w "\n" -X POST http://127.0.0.1:5433/v1/marcua/test.sqlite/query -H "authorization: Bearer <API_TOKEN_FROM_PREVIOUS_COMMAND>" -d "INSERT INTO favorite_databases (name, score) VALUES (\"SQLite\", 9);"

{"fields":[],"rows":[]}

$ curl -w "\n" -X POST http://127.0.0.1:5433/v1/marcua/test.sqlite/query -H "authorization: Bearer <API_TOKEN_FROM_PREVIOUS_COMMAND>" -d "INSERT INTO favorite_databases (name, score) VALUES (\"DuckDB\", 9);"

{"fields":[],"rows":[]}

$ curl -w "\n" -X POST http://127.0.0.1:5433/v1/marcua/test.sqlite/query -H "authorization: Bearer <API_TOKEN_FROM_PREVIOUS_COMMAND>" -d "SELECT * FROM favorite_databases;"

{"fields":["name","score"],"rows":[["PostgreSQL","10"],["SQLite","9"],["DuckDB","9"]]}

Snapshots / backups

You can configure ayb to periodically upload snapshots of each database to S3-compatible storage to recover from the failure of the machine running ayb or revert to a previous copy of the data. Each snapshot is compressed (using zstd) and only uploaded if the database changed since the last snapshot. To enable snapshot-based backups, include a configuration block like the following in your ayb.toml:

[snapshots]
sqlite_method = "Vacuum"
access_key_id = "YOUR_S3_ACCESS_KEY_ID"
secret_access_key = "YOUR_S3_ACCESS_KEY_SECRET"
bucket = "bucket-to-upload-snapshots"
path_prefix = "some/optional/prefix"
endpoint_url = "https://url-endpoint-of-s3-compatible-provider.com"  # Optional
region = "us-east-1"  # Optional
force_path_style = false  # Optional

[snapshots.automation]
interval = "10m"
max_snapshots = 3

Here is an explanation of the parameters:

  • sqlite_method: The two SQLite backup methods are Vacuum and Backup. ayb only supports Vacuum for now.
  • access_key_id / secret_access_key: The access key ID and secret to upload/list snapshots to your S3-compatible storage provider.
  • bucket: The name of the bucket to which to upload snapshots.
  • bucket_prefix: (Can be blank) if you want to upload snapshots to a prefixed path inside bucket (e.g., my-bucket/the-snapshots), provide a prefix (e.g., the-snapshots).
  • endpoint_url: (Optional if using AWS S3) Each S3-compatible storage provider will tell you their own endpoint to manage your buckets.
  • region: (Optional if using AWS S3) Some S3-compatible storage providers will request a region in their network where your bucket will live.
  • force_path_style: (Optional, legacy) If included and true, will use the legacy path-style method of referencing buckets. Used in ayb's end-to-end tests and might be helpful beyond, but start without it.
  • interval: How frequently to take a snapshot of your data in human-readable format (e.g., every 30 minutes = 30m, every hour = 1h, every hour and 30 minutes = 1h30m, with more examples here).
  • max_snapshots: How many old snapshots to keep before pruning the oldest ones.

Once snapshots are enabled, you will see logs on the server with each periodic snapshot run. The following example shows how snapshots work, including how to list and restore them (using interval = "3s" and max_snapshots = 2):

$ ayb client create_database marcua/snapshots.sqlite
Successfully created marcua/snapshots.sqlite

$ ayb client query marcua/snapshots.sqlite "CREATE TABLE favorite_databases(name varchar, score integer);"
Rows: 0

$ ayb client query marcua/snapshots.sqlite "INSERT INTO favorite_databases (name, score) VALUES (\"PostgreSQL\", 10);"
Rows: 0

# Wait longer than 3 seconds before inserting the next row, so that a snapshot with just PostgreSQL exists.
$ ayb client query marcua/snapshots.sqlite "INSERT INTO favorite_databases (name, score) VALUES (\"SQLite\", 9);"
Rows: 0

$ ayb client query marcua/snapshots.sqlite "SELECT * FROM favorite_databases;"
 name       | score
------------+-------
 PostgreSQL | 10
 SQLite     | 9

Rows: 2

# Wait longer than 3 seconds before listing snapshots to ensure that a snapshot with SQLite exists as well.
$ ayb client list_snapshots marcua/snapshots.sqlite
 Name                                                             | Last modified
------------------------------------------------------------------+---------------------------
 f9e01a396fb7f91be988c26d43f9ffa667bd0fd05009b231aa61ea1073d34423 | 2024-08-18T15:05:04+00:00
 856e21f7cae8383426cd2e0599caf6e83962b051af4734ab5c53aff87ea0ff45 | 2024-08-18T15:04:40+00:00

# Restore the older snapshot, which didn't contain SQLite
$ ayb client restore_snapshot marcua/snapshots.sqlite 856e21f7cae8383426cd2e0599caf6e83962b051af4734ab5c53aff87ea0ff45
Restored marcua/snapshots.sqlite to snapshot 856e21f7cae8383426cd2e0599caf6e83962b051af4734ab5c53aff87ea0ff45

$ ayb client query marcua/snapshots.sqlite "SELECT * FROM favorite_databases;"
 name       | score
------------+-------
 PostgreSQL | 10

Rows: 1

Credits: the design of snapshot-based backups was influenced by that of rqlite. Thank you to the authors for their great design and documentation.

Isolation

ayb allows multiple users to run queries against databases that are stored on the same machine. Isolation enables you to prevent one user from accessing another user's data, and allows you to restrict the resources any one user is able to utilize.

By default, ayb uses SQLITE_DBCONFIG_DEFENSIVE flag and sets SQLITE_LIMIT_ATTACHED to 0 in order to prevent users from corrupting the database or attaching to other databases on the filesystem.

For further isolation, ayb uses nsjail to isolate each query's filesystem access and resources. When this form of isolation is enabled, ayb starts a new nsjail-managed process to execute the query against the database. We have not yet benchmarked the performance overhead of this approach.

To enable isolation, you must first build nsjail, which you can do through scripts/build_nsjail.sh. Note that nsjail depends on a few other packages. If you run into issues building it, it might be helpful to see its Dockerfile to get a sense of those requirements.

Once you have a path to the nsjail binary, add the following to your ayb.toml:

[isolation]
nsjail_path = "path/to/nsjail"

Testing

ayb is largely tested through end-to-end tests that mimic as realistic an environment as possible. Individual modules may also provide more specific unit tests. To set up your environment for running end-to-end tests, type:

tests/set_up_e2e_env.sh

After your environment is set up, you can run the tests with:

cargo test --verbose

In order to mimic as close to a realistic environment as possible, the end-to-end tests mock out very little functionality. The tests/set_up_e2e_env.sh script, which has been used extensively in Ubuntu, does the following:

  • Sets up a Python virtual environment and installs requirements for various helpers.
  • Installs requirements for a Python-based stub SMTP server to help test email-based registration.
  • Installs the requirements for a LocalStack server and then runs that server in the background (requires Docker) in order to test database snapshotting functionality that stores snapshots in S3-compatible storage.
  • Installs an nsjail binary to test ayb's isolation functionality.

FAQ

Who is ayb for?

The introductory blog post has a section describing each group that stands to benefit from ayb's aim to make it easier to create a database, interact with it, and share it with relevant people/organizations. Students would benefit from encountering less operational impediments to writing their first SQL query or sharing their in-progress database with a mentor or teacher for help. Sharers like scientists and journalists would benefit from an easy way to post a dataset and share it with collaborators. Finally, anyone concerned about the sovereignty of their data would benefit from a world where it's so easy to spin up a database that more of their data can live in databases they control.

What's with the name?

Thank you for asking. I hope the answer elicits some nostalgia! Shout out to Meelap Shah and Eugene Wu for convincing me to not call this project stacks, to Andrew Lange-Abramowitz for making the connection to the storied meme, and to Meredith Blumenstock for listening to me fret over it all.

Roadmap

Here's a rough roadmap for the project, with items near the top of the list more likely to be completed first. The nitty-gritty list of prioritized issues can be found on this project board, with the most-likely-to-be-completed issues near the top of the to-do list.

  • Make the single-user ayb experience excellent
    • Reduce reliance on PostgreSQL (SQLite metadata storage). Given that the goal of ayb is to make it easier to create, share, and query databases, it's frustrating that running ayb requires you to pay the nontrivial cost of operationalizing PostgreSQL. While Postgres will be helpful for eventually coordinating between multiple ayb nodes, a single-node version should be able to store its metadata in SQLite with little setup costs.
    • Authentication and permissions. Add authentication/the ability to log in, and add permissions to endpoints so that you can't just issue queries against any database.
    • Isolation. Since an ayb instance can have multiple tenants/databases, we want to use one of the many container/isolate/microVM projects to ensure that one tenant isn't able to access another tenant's data.
    • Persistence beyond the node. Back databases up to persistent S3-compatible storage and allow (for now) manual recovery on failure.
    • Clustering. Support for multiple ayb nodes to serve databases and requests. Whereas a single database will not span multiple machines, parallelism/distribution will happen across users and databases.
    • Sessions/transactions. ayb's query API is a stateless request/response API, making it impossible to start a database transaction or issue multiple queries in a session. Exposing sessions in the API will allow multiple statements per session, and by extension, transactions.
    • Import/export of databases. ayb already uses existing well-established file formats (e.g., SQLite). There should be endpoints to import existing databases into ayb in those formats or export the underlying files so you're not locked in.
    • High availablity/automatic failover. While ayb provides snapshot-based backups to protect against cataclysmic failures, the recovery process is manual. Streaming databases to replicas and switching to replicas on failure will make ayb more highly available.
  • Extend ayb to more people and software
    • Collaboration. In addition to making it easy to create and query databases, it should be easy to share databases with others. Two use cases include adding private collaborators and allowing public read-only access.
    • Forking. Allowing a user to fork their own copy of a database will enable collaborators to remix and build on each others' work.
    • Versioning. To both make it less scary to execute sensitive operations and to make it possible for scientists to reference and publish checkpoints of their work, a user should be able to snapshot and revert to a database at a point in time.
    • DuckDB. Allowing users to create a DuckDB database in addition to a SQLite database would allow you to create a data warehouse with a single command. This effort is dependent on the DuckDB project. First, the DuckDB file format is rapidly changing ahead of the project's 1.0 release. Additionally, I don't know of an equivalent streaming replication project to LiteFS for DuckDB that handles persistence beyond the node.
    • PostgreSQL wire protocol. While an HTTP API makes it easy to build new web apps, exposing ayb over the PostgreSQL wire protocol will allow existing tools and libraries to connect to and query an ayb database.
  • Increase discoverability with a web frontend
    • Provide a web interface analogous to the command line interface. Much like GitHub/Gitea/Forgejo make git more approachable, you shouldn't have to pay a command line knowledge tax in order to create, share, and query an ayb database.
    • Explore people's public datasets. Beyond simplifying the command line, platforms like GitHub also make it easier to find a user's publicly shared repositories, follow along in their work, and fork a copy for your own exploration. That same experience should be possible for ayb-hosted databases.

Contributing

(This section is inspired by the LiteFS project, and is just one of the many things to love about that project.)

ayb contributions work a little different than most GitHub projects:

  • If you have a small bug fix or typo fix, please PR directly to this repository.
  • If you want to contribute documentation, please PR directly to this repository.
  • If you would like to contribute a feature, create and discuss the feature in an issue on this GitHub repository first. Once the feature and some of its finer details are hashed out in the issue and potentially a design document, submit a pull request. I might politely decline pull requests that haven't first been discussed/designed.

This project has a roadmap and features are added and tested in a certain order. I'm adding a little friction in requiring a discussion/design document for features before submitting a pull request to ensure that I can focus my attention on well-motivated, well-sequenced, and well-understood functionality.

Dependencies

~79–115MB
~2M SLoC