DefraDB is a user-centric database that prioritizes data ownership, personal privacy, and information security. Its data model, powered by the convergence of MerkleCRDTs and the content-addressability of IPLD, enables a multi-write-master architecture. It features DQL, a query language compatible with GraphQL but providing extra convenience. By leveraging peer-to-peer networking it can be deployed nimbly in novel topologies. Access control is determined by a relationship-based DSL, supporting document or field-level policies, secured by the SourceHub network. DefraDB is a core part of the Source technologies that enable new paradigms of decentralized data and access-control management, user-centric apps, data trustworthiness, and much more.

Read the documentation on docs.source.network.

• Early Access
• Install
• Start
• Configuration
• Add a schema type
• Create a document instance
• Query documents
• Obtain document commits
• DefraDB Query Language (DQL)
• Peer-to-peer data synchronization
  • Pubsub example
  • Collection subscription example
  • Replicator example
• Securing the HTTP API with TLS
• Licensing
• Contributors

DISCLAIMER: At this early stage, DefraDB does not offer access control or data encryption, and the default configuration exposes the database to the network. The software is provided "as is" and is not guaranteed to be stable, secure, or error-free. We encourage you to experiment with DefraDB and provide feedback, but please do not use it for production purposes until it has been thoroughly tested and developed.

Install defradb by downloading an executable or building it locally using the Go toolchain:

git clone [email protected]:sourcenetwork/defradb.git
cd defradb
make install

In the following sections, we assume that defradb is included in your PATH. If you installed it with the Go toolchain, use:

export PATH=$PATH:$(go env GOPATH)/bin

We recommend experimenting with queries using a native GraphQL client. GraphiQL is a popular option - download and install it.

Start a node by executing defradb start. Keep the node running while going through the following examples.

Verify the local connection to the node works by executing defradb client ping in another terminal.

In this document, we use the default configuration, which has the following behavior:

• ~/.defradb/ is DefraDB's configuration and data directory
• client command interacts with the locally running node
• The GraphQL endpoint is provided at https://localhost:9181/api/v0/graphql

The GraphQL endpoint can be used with a GraphQL client (e.g., Altair) to conveniently perform requests (query, mutation) and obtain schema introspection.

Schemas are used to structure documents using a type system.

In the following examples, we'll be using a simple User schema type.

Add it to the database with the following command. By doing so, DefraDB generates the typed GraphQL endpoints for querying, mutation, and introspection.

defradb client schema add '
  type User {
    name: String 
    age: Int 
    verified: Boolean 
    points: Float
  }
'

Find more examples of schema type definitions in the examples/schema/ folder.

Submit a mutation request to create an docuement of the User type:

defradb client query '
  mutation {
      create_User(data: "{\"age\": 31, \"verified\": true, \"points\": 90, \"name\": \"Bob\"}") {
          _key
      }
  }
'

Expected response:

{
  "data": [
    {
      "_key": "bae-91171025-ed21-50e3-b0dc-e31bccdfa1ab",
    }
  ]
}

_key is the document's key, a unique identifier of the document, determined by its schema and initial data.

Once you have populated your node with data, you can query it:

defradb client query '
  query {
    User {
      _key
      age
      name
      points
    }
  }
'

This query obtains all users and returns their fields _key, age, name, points. GraphQL queries only return the exact fields requested.

You can further filter results with the filter argument.

defradb client query '
  query {
    User(filter: {points: {_ge: 50}}) {
      _key
      age
      name
      points
    }
  }
'

This returns only user documents which have a value for the points field Greater Than or Equal to (_ge) 50.

DefraDB's data model is based on MerkleCRDTs. Each document has a graph of all of its updates, similar to Git. The updates are called commits and are identified by cid, a content identifier. Each references its parents by their cids.

To get the most recent commit in the MerkleDAG for the document identified as bae-91171025-ed21-50e3-b0dc-e31bccdfa1ab:

defradb client query '
  query {
    latestCommits(dockey: "bae-91171025-ed21-50e3-b0dc-e31bccdfa1ab") {
      cid
      delta
      height
      links {
        cid
        name
      }
    }
  }
'

It returns a structure similar to the following, which contains the update payload that caused this new commit (delta) and any subgraph commits it references.

{
  "data": [
    {
      "cid": "bafybeifhtfs6vgu7cwbhkojneh7gghwwinh5xzmf7nqkqqdebw5rqino7u",
      "delta": "pGNhZ2UYH2RuYW1lY0JvYmZwb2ludHMYWmh2ZXJpZmllZPU=",
      "height": 1,
      "links": [
        {
          "cid": "bafybeiet6foxcipesjurdqi4zpsgsiok5znqgw4oa5poef6qtiby5hlpzy",
          "name": "age"
        },
        {
          "cid": "bafybeielahxy3r3ulykwoi5qalvkluojta4jlg6eyxvt7lbon3yd6ignby",
          "name": "name"
        },
        {
          "cid": "bafybeia3tkpz52s3nx4uqadbm7t5tir6gagkvjkgipmxs2xcyzlkf4y4dm",
          "name": "points"
        },
        {
          "cid": "bafybeia4off4javopmxcdyvr6fgb5clo7m5bblxic5sqr2vd52s6khyksm",
          "name": "verified"
        }
      ]
    }
  ]
}

Obtain a specific commit by its content identifier (cid):

defradb client query '
  query {
    commits(cid: "bafybeifhtfs6vgu7cwbhkojneh7gghwwinh5xzmf7nqkqqdebw5rqino7u") {
      cid
      delta
      height
      links {
        cid
        name
      }
    }
  }
'

DQL is compatible with GraphQL but features various extensions.

Read its documentation at docs.source.network to discover its filtering, ordering, limiting, relationships, variables, aggregate functions, and other useful features.

DefraDB leverages peer-to-peer networking for data exchange, synchronization, and replication of documents and commits.

When starting a node for the first time, a key pair is generated and stored in its "root directory" (~/.defradb/ by default).

Each node has a unique PeerID generated from its public key. This ID allows other nodes to connect to it.

There are two types of peer-to-peer relationships supported: pubsub peering and replicator peering.

Pubsub peering passively synchronizes data between nodes by broadcasting Document Commit updates to the topic of the commit's document key. Nodes need to be listening on the pubsub channel to receive updates. This is for when two nodes already have share a document and want to keep them in sync.

Replicator peering actively pushes changes from a specific collection to a target peer.

Pubsub peers can be specified on the command line using the --peers flag, which accepts a comma-separated list of peer multiaddresses. For example, a node at IP 192.168.1.12 listening on 9000 with PeerID 12D3KooWNXm3dmrwCYSxGoRUyZstaKYiHPdt8uZH5vgVaEJyzU8B would be referred to using the multiaddress /ip4/192.168.1.12/tcp/9000/p2p/12D3KooWNXm3dmrwCYSxGoRUyZstaKYiHPdt8uZH5vgVaEJyzU8B.

Let's go through an example of two nodes (nodeA and nodeB) connecting with each other over pubsub, on the same machine.

Start nodeA with a default configuration:

defradb start

Obtain the PeerID from its console output. In this example, we use 12D3KooWNXm3dmrwCYSxGoRUyZstaKYiHPdt8uZH5vgVaEJyzU8B, but locally it will be different.

For nodeB, we provide the following configuration:

defradb start --rootdir ~/.defradb-nodeB --url localhost:9182 --p2paddr /ip4/0.0.0.0/tcp/9172 --tcpaddr /ip4/0.0.0.0/tcp/9162 --peers /ip4/0.0.0.0/tcp/9171/p2p/12D3KooWNXm3dmrwCYSxGoRUyZstaKYiHPdt8uZH5vgVaEJyzU8B

About the flags:

• --rootdir specifies the root dir (config and data) to use
• --url is the address to listen on for the client HTTP and GraphQL API
• --p2paddr is the multiaddress for the p2p networking to listen on
• --tcpaddr is the multiaddress for the gRPC server to listen on
• --peers is a comma-separated list of peer multiaddresses

This starts two nodes and connects them via pubsub networking.

It is possible to subscribe to updates on a given collection by using its ID as the pubsub topic. The ID of a collection is found as the field schemaVersionID in one of its documents. Here we use the collection ID of the User type we created above. After setting up 2 nodes as shown in the Pubsub example section, we can subscribe to collections updates on nodeA from nodeB by using the rpc p2pcollection command:

defradb client rpc p2pcollection add --url localhost:9182 bafkreibpnvkvjqvg4skzlijka5xe63zeu74ivcjwd76q7yi65jdhwqhske

Multiple collection IDs can be added at once.

defradb client rpc p2pcollection add --url localhost:9182 <collection1ID> <collection2ID> <collection3ID>

Replicator peering is targeted: it allows a node to actively send updates to another node. Let's go through an example of nodeA actively replicating to nodeB:

Start nodeA:

defradb start

In another terminal, add this example schema to it:

defradb client schema add '
  type Article {
    content: String
    published: Boolean
  }
'

Start (or continue running from above) nodeB, that will be receiving updates:

defradb start --rootdir ~/.defradb-nodeB --url localhost:9182 --p2paddr /ip4/0.0.0.0/tcp/9172 --tcpaddr /ip4/0.0.0.0/tcp/9162

Here we do not specify --peers as we will manually define a replicator after startup via the rpc client command.

In another terminal, add the same schema to nodeB:

defradb client schema add --url localhost:9182 '
  type Article {
    content: String
    published: Boolean
  }
'

Set nodeA to actively replicate the "Article" collection to nodeB:

defradb client rpc replicator set -c "Article" /ip4/0.0.0.0/tcp/9172/p2p/<peerID_of_nodeB>

As we add or update documents in the "Article" collection on nodeA, they will be actively pushed to nodeB. Note that changes to nodeB will still be passively published back to nodeA, via pubsub.

By default, DefraDB will expose its HTTP API at https://localhost:9181/api/v0. It's also possible to configure the API to use TLS with self-signed certificates or Let's Encrypt.

To start defradb with self-signed certificates placed under ~/.defradb/certs/ with server.key being the public key and server.crt being the private key, just do:

defradb start --tls

The keys can be generated with your generator of choice or with make tls-certs.

Since the keys should be stored within the DefraDB data and configuration directory, the recommended key generation command is make tls-certs path="~/.defradb/certs".

If not saved under ~/.defradb/certs then the public (pubkeypath) and private (privkeypaths) key paths need to be explicitly defined in addition to the --tls flag or tls set to true in the config.

Then to start the server with TLS, using your generated keys in custom path:

defradb start --tls --pubkeypath ~/path-to-pubkey.key --privkeypath ~/path-to-privkey.crt

DefraDB also comes with automatic HTTPS for deployments on the public web. To enable HTTPS, deploy DefraDB to a server with both port 80 and port 443 open. With your domain's DNS A record pointed to the IP of your server, you can run the database using the following command:

sudo defradb start --tls --url=your-domain.net [email protected]

Note: sudo is needed above for the redirection server (to bind port 80).

A valid email address is necessary for the creation of the certificate, and is important to get notifications from the Certificate Authority - in case the certificate is about to expire, etc.

When accessing DefraDB through a frontend interface, you may be confronted with a CORS error. That is because, by default, DefraDB will not have any allowed origins set. To specify which origins should be allowed to access your DefraDB endpoint, you can specify them when starting the database:

defradb start --allowed-origins=https://yourdomain.com

If running a frontend app locally on localhost, allowed origins must be set with the port of the app:

defradb start --allowed-origins=https://localhost:3000

The catch-all * is also a valid origin.

Discuss on Discord or Github Discussions. The Source project is on Twitter.

DefraDB's code is released under the Business Source License (BSL). It grants you the right to copy, modify, create derivative works, redistribute, and make non-production use of it. For additional uses, such as deploying in production on a private network, please contact [email protected] for a licensing agreement. Each dated version of the license turns into the more permissive Apache License v2.0 after four years. Please read the complete license before usage.

• John-Alan Simmons (@jsimnz)
• Andrew Sisley (@AndrewSisley)
• Shahzad Lone (@shahzadlone)
• Orpheus Lummis (@orpheuslummis)
• Fred Carle (@fredcarle)
• Islam Aliev (@islamaliev)

You are invited to contribute to DefraDB. Follow the Contributing guide to get started.