NDK (Nostr Development Kit) is a Dart library designed to enhance the Nostr development experience.\ It provides streamlined solutions for common use cases and abstracts away complex relay management, making it ideal for building constrained Nostr clients, particularly on mobile devices.\ NDK implements the inbox/outbox (gossip) model by default, optimizing network usage and improving performance.
Table of Contents:
$~~~$
Rust toolchain android:
rustup target add \
aarch64-linux-android \
armv7-linux-androideabi \
x86_64-linux-android \
i686-linux-android
Rust toolchain ios:
# 64 bit targets (real device & simulator):
rustup target add aarch64-apple-ios x86_64-apple-ios
# New simulator target for Xcode 12 and later
rustup target add aarch64-apple-ios-sim
# 32 bit targets (you probably don't need these):
rustup target add armv7-apple-ios i386-apple-ios
flutter pub add ndk
import 'package:ndk/ndk.dart';
usage examples
import 'package:ndk/ndk.dart';
// init
Ndk ndk = Ndk(
NdkConfig(
eventVerifier: RustEventVerifier(),
cache: MemCacheManager(),
),
);
// query
NdkResponse response = ndk.requests.query(
filters: [
Filter(
authors: ['hexPubkey']
kinds: [Nip01Event.TEXT_NODE_KIND],
limit: 10,
),
],
);
// result
await for (final event in response.stream) {
print(event);
}
$~~~$
inMemory
bip340, rust
bip340, amber
There are two main constrains that we aim for: battery/compute and network bandwidth.
network\ Inbox/Outbox (gossip) is our main pillar to help avoid unnecessary nostr requests. We try to leverage the cache as much as possible. \ Even splitting the users filters into smaller relay tailored filters if we know the relay has the information we need.
compute\
Right now the most compute intensive operation is verifying signatures. \
We use the cache to determine if we have already seen the event and only if it is unknown signature verification is done. \
To make the operation as optimized as possible we strongly recommend using RustEventVerifier()
because it uses a separate thread for verification.
$~~~$
The simplest characterization of the gossip model is just this: reading the posts of people you follow from the relays that they wrote them to.
more details on https://mikedilger.com/gossip-model/
term | explanation | simmilar to |
---|---|---|
broadcastEvent | push event to nostr network/relays | postEvent, publishEvent |
JIT | Just In Time, e.g. as it happens | - |
query | get data once and close the request | get request |
subscription | stream of events as they come in | stream of data |
bootstrapRelays | default relays to connect when nothing else is specified | seed relays, initial relays |
engine | optimized network resolver for nostr requests | - |
$~~~$
$~~~$
Install prerequisites
If you work on rust code (rust_builder/rust
) run flutter_rust_bridge_codegen generate --watch
to generate the rust dart glue code.
Run build runner: (e.g for generating mocks)\
dart run build_runner build
This project uses Clean Architecture. Reasons for it being clear separation of concerns and therefore making it more accessible for future contributors.\ You can read more about it here.
For initialization we use presentation_layer/init.dart
to assemble all dependencies, these are then exposed in presentation_layer/ndk.dart
the main entry point for the lib user.
Global state is realized via a simple [GlobalState] object created by ndk.dart
. \
The lib user is supposed to keep the [NDK] object in memory.
Other state objects are created on demand, for example [RequestState] for each request.
lib/
├── config/
│ └── # Configuration files
├── shared/
│ ├── nipX/ # NIP-specific code folders
│ └── # Internal code, no external dependencies
├── data_layer/
│ ├── data_sources/
│ │ └── # External APIs, WebSocket implementations, etc.
│ ├── models/
│ │ └── # Data transformation (e.g., JSON to entity)
│ └── repositories/
│ └── # Concrete repository implementations
├── domain_layer/
│ ├── entities/
│ │ └── # Core business objects
│ ├── repositories/
│ │ └── # Repository contracts
│ └── usecases/
│ └── # Business logic / use cases
├── presentation_layer/
│ └── # API design (exposing use cases to the outside world)
└── ndk.dart # Entry point, directs to presentation layer
NDK ships with two network Engines. An Engine is part of the code that resolves nostr requests over the network and handles the WebSocket connections.\ Its used to handle the inbox/outbox (gossip) model efficiently.
Lists Engine:\ Precalculates the best possible relays based on nip65 data. During calculation relay connectivity is taken into account. This works by connecting and checking the health status of a relay before its added to the ranking pool.\ This method gets close to the optimal connections given a certain pubkey coverage.
Just in Time (JIT) Engine:\ JIT Engine does the ranking on the fly only for the missing coverage/pubkey. Healthy relays are assumed during ranking and replaced later on if a relay fails to connect.\ To Avoid rarely used relays and spawning a bunch of unessecary connections, already connected relays get a boost, and a usefulness score is considered for the ranking.\ For more information look here
Custom Engine\ If you want to implement your own engine with custom behavior you need to touch the following things:
NetworkEngine
interfacenetworkController
in the RequestState
global_state.dart
init.dart
The current state solution is not ideal because it requires coordination between the engine authors and not enforceable by code. If you have ideas how to improve this system, please reach out.
The network engine is only concerned about network requests! Caching and avoiding concurrency is handled by separate usecases. Take a look at
requests.dart
usecase to learn more.