• Stars
    star
    1,430
  • Rank 32,678 (Top 0.7 %)
  • Language
    Swift
  • License
    Apache License 2.0
  • Created almost 6 years ago
  • Updated 8 months ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

A light-weight server-side service framework written in the Swift programming language.

Build - Main Branch Swift 5.6, 5.7 and 5.8 Tested Ubuntu 20.04 and 22.04 Tested CentOS 8 Tested Amazon Linux 2 Tested Join the Smoke Server Side community on gitter Apache 2

Smoke Framework

The Smoke Framework is a light-weight server-side service framework written in Swift and using SwiftNIO for its networking layer by default. The framework can be used for REST-like or RPC-like services and in conjunction with code generators from service models such as Swagger/OpenAPI.

The framework has built in support for JSON-encoded request and response payloads.

Support Policy

SmokeFramework follows the same support policy as followed by SmokeAWS here.

Conceptual Overview

The Smoke Framework provides the ability to specify handlers for operations your service application needs to perform. When a request is received, the framework will decode the request into the operation's input. When the handler returns, its response (if any) will be encoded and sent in the response.

Each invocation of a handler is also passed an application-specific context, allowing application-scope or invocation-scope entities such as other service clients to be passed to operation handlers. Using the context allows operation handlers to remain pure functions (where its return value is determined by the function's logic and input values) and hence easily testable.

SmokeFrameworkExamples

See this repository for examples of the Smoke Framework and the related Smoke* repositories in action.

Getting Started using Code Generation

The Smoke Framework provides a code generator that will generate a complete Swift Package Manager repository for a SmokeFramework-based service from a Swagger 2.0 specification file.

See the instructions in the code generator repository on how to get started.

Getting Started without Code Generation

These steps assume you have just created a new swift application using swift package init --type executable.

Step 1: Add the Smoke Framework dependency

The Smoke Framework uses the Swift Package Manager. To use the framework, add the following dependency to your Package.swift-

dependencies: [
    .package(url: "https://github.com/amzn/smoke-framework.git", from: "2.0.0")
]

.target(name: ..., dependencies: [
    ..., 
    .product(name: "SmokeOperationsHTTP1Server", package: "smoke-framework"),
]),

Step 2: Update the runtime dependency requirements of the application

If you attempt to compile the application, you will get the error

the product 'XXX' requires minimum platform version 10.12 for macos platform

This is because SmokeFramework projects have a minimum MacOS version dependency. To correct this there needs to be a couple of additions to to the Package.swift file.

Step 2a: Update the language version

Specify the language versions supported by the application-

targets: [
    ...
    ],
swiftLanguageVersions: [.v5]

Step 2b: Update the supported platforms

Specify the platforms supported by the application-

name: "XXX",
platforms: [
  .macOS(.v10_15), .iOS(.v10)
],
products: [

Step 2: Add a Context Type

An instance of the context type will be passed to each invocation of an operation that needs to be handled. This instance can be setup to be initialized per invocation or once for the application.

You will need to create this context type. There are no requirements for a type to be passed as a context. The following code shows an example of creating the context type-

public struct MyApplicationContext {
    let logger: Logger
    // TODO: Add properties to be accessed by the operation handlers
    
    public init(logger: Logger) {
        self.logger = logger
    }
}

Step 3: Add an Operation Function

The next step to using the Smoke Framework is to define one or more functions that will perform the operations that your application requires. The following code shows an example of such a function-

extension MyApplicationContext {
    func handleTheOperation(input: OperationInput) throws -> OperationOutput {
        return OperationOutput()
    }
}

This particular operation function accepts the input to the operation and is within an extension of the context (giving it access to any attributes or functions on this type) while returning the output from the operation.

For HTTP1, the operation input can conform to OperationHTTP1InputProtocol, which defines how the input type is constructed from the HTTP1 request. Similarly, the operation output can conform to OperationHTTP1OutputProtocol, which defines how to construct the HTTP1 response from the output type. Both must also conform to the Validatable protocol, giving the opportunity to validate any field constraints.

As an alternative, both operation input and output can conform to the Codable protocol if they are constructed from only one part of the HTTP1 request and response.

The Smoke Framework also supports additional built-in and custom operation function signatures. See the The Operation Function and Extension Points sections for more information.

Step 4: Add Handler Selection

After defining the required operation handlers, it is time to specify how they are selected for incoming requests.

The Smoke Framework provides the SmokeHTTP1HandlerSelector protocol to add handlers to a selector.

import SmokeOperationsHTTP1

public enum MyOperations: String, Hashable, CustomStringConvertible {
    case theOperation = "TheOperation"

    public var description: String {
        return rawValue
    }

    public var operationPath: String {
        switch self {
        case .theOperation:
            return "/theOperation"
        }
    }
}

public extension MyOperations {
    static func addToSmokeServer<SelectorType: SmokeHTTP1HandlerSelector>(selector: inout SelectorType)
            where SelectorType.ContextType == MyApplicationContext,
                  SelectorType.OperationIdentifer == MyOperations {
        
        let allowedErrorsForTheOperation: [(MyApplicationErrors, Int)] = [(.unknownResource, 404)]
        selector.addHandlerForOperationProvider(.theOperation, httpMethod: .POST,
                                                operationProvider: MyApplicationContext.handleTheOperation,
                                                allowedErrors: allowedErrorsForTheOperation)
    }
}

Each handler added requires the following parameters to be specified:

  • The operation to be added. This must be of a type conforming to OperationIdentity such as an enum.
    • The HTTP method that must be matched by the incoming request to select the handler.
    • The function to be invoked.
    • The errors that can be returned to the caller from this handler. The error type must also conform to CustomStringConvertible that returns the identity of the current error.
    • The location in the HTTP1 request to construct the operation input type from (only required if the input type conforms to Codable)
    • The location in the HTTP1 response that the output type represents (only required if the output type conforms to Codable)

Step 5: Setting up the Application Server

The final step is to setup an application as an operation server.

import Foundation
import SmokeOperationsHTTP1Server
import AsyncHTTPClient
import NIO
import SmokeHTTP1

struct MyPerInvocationContextInitializer: StandardJSONSmokeServerPerInvocationContextInitializer {
    typealias ContextType = MyApplicationContext
    typealias OperationIdentifer = MyOperations
    
    let serverName = "MyService"
    // specify the operations initializer
    let operationsInitializer: OperationsInitializerType = MyOperations.addToSmokeServer

    /**
     On application startup.
     */
    init(eventLoopGroup: EventLoopGroup) throws {
        // set up any of the application-wide context
    }

    /**
     On invocation.
    */
    public func getInvocationContext(
        invocationReporting: SmokeServerInvocationReporting<SmokeInvocationTraceContext>) -> MyApplicationContext {
        // create an invocation-specific context to be passed to an operation handler
        return MyApplicationContext(logger: invocationReporting.logger)
    }

    /**
     On application shutdown.
    */
    func onShutdown() throws {
        // shutdown anything before the application closes
    }
}

SmokeHTTP1Server.runAsOperationServer(MyPerInvocationContextInitializer.init)

You can now run the application and the server will start up on port 8080. The application will block in the SmokeHTTP1Server.runAsOperationServer call. When the server has been fully shutdown and has completed all requests, onShutdown will be called. In this function you can close/shutdown any clients or credentials that were created on application startup.

Step 6: Add Reporting Configuration (Optional)

An optional configuration step is to setup the reporting configuration for metrics emitted by the Smoke Framework. This involves overriding the default reportingConfiguration attribute on the initializer. For the metrics to be emitted, a swift-metrics backend - such as CloudWatchMetricsFactory - will need to be initialized.

...

struct MyPerInvocationContextInitializer: StandardJSONSmokeServerPerInvocationContextInitializer {
    typealias ContextType = MyApplicationContext
    typealias OperationIdentifer = MyOperations
    
    let reportingConfiguration: SmokeReportingConfiguration<OperationIdentifer>
    let serverName = "MyService"
    // specify the operations initializer
    let operationsInitializer: OperationsInitializerType = MyOperations.addToSmokeServer

    /**
     On application startup.
     */
    init(eventLoopGroup: EventLoopGroup) throws {
        // set up any of the application-wide context
        
        // for the server, only report the latency metrics
        // only report 5XX error counts for TheOperation (even if additional operations are added in the future)
        // only report 4XX error counts for operations other than TheOperation (as they are added in the future)
        self.reportingConfiguration = SmokeReportingConfiguration(
            successCounterMatchingRequests: .none,
            failure5XXCounterMatchingRequests: .onlyForOperations([.theOperation]),
            failure4XXCounterMatchingRequests: .exceptForOperations([.theOperation]),
            requestReadLatencyTimerMatchingRequests: .none,
            latencyTimerMatchingRequests: .all,
            serviceLatencyTimerMatchingRequests: .all,
            outwardServiceCallLatencyTimerMatchingRequests: .all,
            outwardServiceCallRetryWaitTimerMatchingRequests: .all)
    }

    ...
}

SmokeHTTP1Server.runAsOperationServer(MyPerInvocationContextInitializer.init)

Enabling Distributed Tracing (Swift 5.7 and greater)

To have your application participate in distributed traces, add a property enableTracingWithSwiftConcurrency with a value of true to your application initializer.

struct MyPerInvocationContextInitializer: StandardJSONSmokeServerPerInvocationContextInitializer {
    let enableTracingWithSwiftConcurrency = true

    ...
}

This will enable tracing for any operation handlers that use Swift Concurrency (async/await). You will also need to setup an Instrumentation backend by following the instructions here.

Further Concepts

The Application Context

An instance of the application context type is created at application start-up and is passed to each invocation of an operation handler. The framework imposes no restrictions on this type and simply passes it through to the operation handlers. It is recommended that this context is immutable as it can potentially be passed to multiple handlers simultaneously. Otherwise, the context type is responsible for handling its own thread safety.

It is recommended that applications use a strongly typed context rather than a bag of stuff such as a Dictionary.

The Operation Delegate

The Operation Delegate handles specifics such as encoding and decoding requests to the handler's input and output.

The Smoke Framework provides the JSONPayloadHTTP1OperationDelegate implementation that expects a JSON encoded request body as the handler's input and returns the output as the JSON encoded response body.

Each addHandlerForOperation invocation can optionally accept an operation delegate to use when that handler is selected. This can be used when operations have specific encoding or decoding requirements. A default operation delegate is set up at server startup to be used for operations without a specific handler or when no handler matches a request.

The Trace Context

The JSONPayloadHTTP1OperationDelegate takes a generic parameter conforming to the HTTP1OperationTraceContext protocol. This protocol can be used to providing request-level tracing. The requirements for this protocol are defined here.

A default implementation - SmokeInvocationTraceContext - provides some basic tracing using request and response headers.

The Operation Function

Each handler provides a function to be invoked when the handler is selected. By default, the Smoke framework provides four function signatures declared on the Context Type that this function can conform to-

  • ((InputType) throws -> ()): Synchronous method with no output.
  • ((InputType) throws -> OutputType): Synchronous method with output.
  • ((InputType, (Swift.Error?) -> ()) throws -> ()): Asynchronous method with no output.
  • ((InputType, (Result<OutputType, Error>) -> ()) throws -> ()): Asynchronous method with output.

Due to Swift type inference, a handler can switch between these different signatures without changing the handler selector declaration - simply changing the function signature is sufficient.

The synchronous variants will return a response as soon as the function returns either with an empty body or the encoded return value. The asynchronous variants will return a response when the provided result handlers are called.

public protocol Validatable {
    func validate() throws
}

In all cases, the InputType and OutputType types must conform to the Validatable protocol. This protocol gives a type the opportunity to verify its fields - such as for string length, numeric range validation. The Smoke Framework will call validate on operation inputs before passing it to the handler and operation outputs after receiving from the handler-

  • If an operation input fails its validation call (by throwing an error), the framework will fail the operation with a 400 ValidationError response, indicating an error by the caller (the framework also logs this event at Info level).

  • If an operation output fails its validation call (by throwing an error), the framework will fail the operation with a 500 Internal Server Error, indicating an error by the service logic (the framework also logs this event at Error level).

    Additionally, the Smoke Framework provides the option to declare operation handlers outside the Context Type as standalone functions. The Context Type is passed in directly to these functions.

    func handleTheOperation(input: OperationInput, context: MyApplicationContext) throws -> OperationOutput {
        return OperationOutput()
    }
    

    Adding the handler selection is slightly different in this case-

    import SmokeOperationsHTTP1
    
    public func addOperations<SelectorType: SmokeHTTP1HandlerSelector>(selector: inout SelectorType)
            where SelectorType.ContextType == MyApplicationContext,
                  SelectorType.OperationIdentifer == MyOperations {
        
        let allowedErrorsForTheOperation: [(MyApplicationErrors, Int)] = [(.unknownResource, 404)]
        selector.addHandlerForOperation(.theOperation, httpMethod: .POST,
                                        operation: handleTheOperation,
                                        allowedErrors: allowedErrorsForTheOperation)
    }
    

The four function signatures are also available when using this style of operation handlers.

  • ((InputType, ContextType) throws -> ()): Synchronous method with no output.
  • ((InputType, ContextType) throws -> OutputType): Synchronous method with output.
  • ((InputType, ContextType, (Swift.Error?) -> ()) throws -> ()): Asynchronous method with no output.
  • ((InputType, ContextType, (Result<OutputType, Error>) -> ()) throws -> ()): Asynchronous method with output.

Error Handling

By default, any errors thrown from an operation handler will fail the operation and the framework will return a 500 Internal Server Error to the caller (the framework also logs this event at Error level). This behavior prevents any unintentional leakage of internal error information.

public typealias ErrorIdentifiableByDescription = Swift.Error & CustomStringConvertible
public typealias SmokeReturnableError = ErrorIdentifiableByDescription & Encodable

Errors can be explicitly encoded and returned to the caller by conforming to the Swift.Error, CustomStringConvertible and Encodable protocols and being specified under allowedErrors in the addHandlerForUri call setting up the operation handler. For example-

public enum MyError: Swift.Error {
    case theError(reason: String)
    
    enum CodingKeys: String, CodingKey {
        case reason = "Reason"
    }
}

extension MyError: Encodable {
    public func encode(to encoder: Encoder) throws {
        var container = encoder.container(keyedBy: CodingKeys.self)
        
        switch self {
        case .theError(reason: let reason):
            try container.encode(reason, forKey: .reason)
        }
    }
}

extension MyError: CustomStringConvertible {
    public var description: String {
        return "TheError"
    }
}

When such an error is returned from an operation handler-

  • A response is returned to the caller with the HTTP code specified in the allowedErrors entry with a payload of the error encoded according to the Encodable protocol.
  • In addition, the provided error identity of the error will be specified in the __type field of the returned payload.
  • Comparison between the error specified in the allowedErrors list and the error thrown from the operation handler is a string comparison between the respective error identities. This is to allow equivalent errors of differing type (such as code generated errors from different models) to be handled as the same error.
  • For the built-in asynchronous operation functions, errors can either be thrown synchronously from the function itself or passed asynchronously to the result handler. Either way, the operation will fail according to the type of error thrown or passed. This is to avoid functions having to catch synchronous errors (such as in setup) only to pass them to the result handler.

Testing

The Smoke Framework has been designed to make testing of operation handlers straightforward. It is recommended that operation handlers are pure functions (where its return value is determined by the function's logic and input values). In this case, the function can be called in unit tests with appropriately constructed input and context instances.

It is recommended that the application-specific context be used to vary behavior between release and testing executions - such as mocking service clients, random number generators, etc. In general this will create more maintainable tests by keeping all the testing logic in the testing function.

If you want to run all test cases in Smoke Framework, please open command line and go to smoke-framework (root) directory, run swift test and then you should be able to see test cases result.

Extension Points

The Smoke Framework is designed to be extensible beyond its current functionality-

  • JSONPayloadHTTP1OperationDelegate provides basic JSON payload encoding and decoding. Instead, the HTTP1OperationDelegate protocol can be used to create a delegate that provides alternative payload encoding and decoding. Instances of this protocol are given the entire HttpRequestHead and request body when decoding the input and encoding the output for situations when these are required.
  • StandardSmokeHTTP1HandlerSelector provides a handler selector that compares the HTTP URI and verb to select a handler. Instead, the SmokeHTTP1HandlerSelector protocol can be used to create a selector that can use any property from the HTTPRequestHead (such as headers) to select a handler.
  • Even if StandardSmokeHTTP1HandlerSelector does fit your requirements, it can be extended to support additional function signatures. See the built-in function signatures (one can be found in OperationHandler+nonblockingWithInputWithOutput.swift) for examples of this.
  • The Smoke Framework currently supports HTTP1 but can be extended to additional protocols while using the same operation handlers if needed. The initializers of OperationHandler provide a protocol-agnostic layer - as an example [1] - which can be used by a protocol-specific layer - such as [2] for HTTP1 - to abstract protocol-specific handling for the different operation types.

[1] https://github.com/amzn/smoke-framework/blob/master/Sources/SmokeOperations/OperationHandler%2BblockingWithInputWithOutput.swift

[2] https://github.com/amzn/smoke-framework/blob/master/Sources/SmokeOperationsHTTP1/SmokeHTTP1HandlerSelector%2BblockingWithInputWithOutput.swift

License

This library is licensed under the Apache 2.0 License.

More Repositories

1

style-dictionary

A build system for creating cross-platform styles.
JavaScript
3,855
star
2

computer-vision-basics-in-microsoft-excel

Computer Vision Basics in Microsoft Excel (using just formulas)
2,391
star
3

selling-partner-api-docs

This repository contains documentation for developers to use to call Selling Partner APIs.
1,541
star
4

alexa-skills-kit-js

SDK and example code for building voice-enabled skills for the Amazon Echo.
1,134
star
5

ion-java

Java streaming parser/serializer for Ion.
Java
840
star
6

sketch-constructor

Read/write/manipulate Sketch files in Node without Sketch plugins!
JavaScript
538
star
7

selling-partner-api-models

This repository contains OpenAPI models for developers to use when developing software to call Selling Partner APIs.
Mustache
532
star
8

pecos

PECOS - Prediction for Enormous and Correlated Spaces
Python
501
star
9

amzn-drivers

Official AWS drivers repository for Elastic Network Adapter (ENA) and Elastic Fabric Adapter (EFA)
C
444
star
10

ion-js

A JavaScript implementation of Amazon Ion.
TypeScript
323
star
11

convolutional-handwriting-gan

ScrabbleGAN: Semi-Supervised Varying Length Handwritten Text Generation (CVPR20)
Python
260
star
12

xfer

Transfer Learning library for Deep Neural Networks.
Python
250
star
13

awsssmchaosrunner

Amazon's light-weight library for chaos engineering on AWS. It can be used for EC2 and ECS (with EC2 launch type).
Kotlin
247
star
14

ion-python

A Python implementation of Amazon Ion.
Python
210
star
15

amazon-pay-sdk-php

Amazon Pay PHP SDK
PHP
209
star
16

fire-app-builder

Fire App Builder is a framework for building java media apps for Fire TV, allowing you to add your feed of media content to a configuration file and build an app to browse and play it quickly.
Java
178
star
17

exoplayer-amazon-port

Official port of ExoPlayer for Amazon devices
Java
168
star
18

oss-dashboard

A dashboard for viewing many GitHub organizations at once.
Ruby
158
star
19

ion-c

A C implementation of Amazon Ion.
C
149
star
20

metalearn-leap

Original PyTorch implementation of the Leap meta-learner (https://arxiv.org/abs/1812.01054) along with code for running the Omniglot experiment presented in the paper.
Python
147
star
21

ion-go

A Go implementation of Amazon Ion.
Go
146
star
22

distance-assistant

Pedestrian monitor that provides visual feedback to help ensure proper social distancing guidelines are being observed
Python
135
star
23

auction-gym

AuctionGym is a simulation environment that enables reproducible evaluation of bandit and reinforcement learning methods for online advertising auctions.
Jupyter Notebook
135
star
24

hawktracer

HawkTracer is a highly portable, low-overhead, configurable profiling tool built in Amazon Video for getting performance metrics from low-end devices.
C++
131
star
25

trans-encoder

Trans-Encoder: Unsupervised sentence-pair modelling through self- and mutual-distillations
Python
131
star
26

smoke-aws

AWS services integration for the Smoke Framework
Swift
109
star
27

amazon-payments-magento-2-plugin

Extension to enable Amazon Pay on Magento 2
PHP
105
star
28

MXFusion

Modular Probabilistic Programming on MXNet
Python
102
star
29

amazon-weak-ner-needle

Named Entity Recognition with Small Strongly Labeled and Large Weakly Labeled Data
Python
99
star
30

amazon-advertising-api-php-sdk

⛔️ DEPRECATED - Amazon Advertising API PHP Client Library
PHP
93
star
31

ion-rust

Rust implementation of Amazon Ion
Rust
86
star
32

ads-advanced-tools-docs

Code samples and supplements for the Amazon Ads advanced tools center
Jupyter Notebook
83
star
33

image-to-recipe-transformers

Code for CVPR 2021 paper: Revamping Cross-Modal Recipe Retrieval with Hierarchical Transformers and Self-supervised Learning
Python
82
star
34

oss-attribution-builder

The OSS Attribution Builder is a website that helps teams create attribution documents (notices, "open source screens", credits, etc) commonly found in software products.
TypeScript
79
star
35

smoke-http

Specialised HTTP Client for service operations abstracted from the HTTP protocol.
Swift
69
star
36

amazon-ray

Staging area for ongoing enhancements to Ray focused on improving integration with AWS and other Amazon technologies.
Python
66
star
37

alexa-coho

Sample code for building skill adapters for Alexa Connected Home using the Lighting API
JavaScript
62
star
38

amazon-pay-sdk-ruby

Amazon Pay Ruby SDK
Ruby
58
star
39

amazon-pay-sdk-java

Amazon Pay Java SDK
Java
53
star
40

amazon-pay-sdk-python

Amazon Pay Python SDK
Python
53
star
41

zero-shot-rlhr

Python
51
star
42

supply-chain-simulation-environment

Python
49
star
43

amazon-pay-sdk-csharp

Amazon Pay C# SDK
C#
47
star
44

ion-dotnet

A .NET implementation of Amazon Ion.
C#
47
star
45

multiconer-baseline

Python
47
star
46

amazon-pay-api-sdk-php

Amazon Pay API SDK (PHP)
PHP
47
star
47

zeek-plugin-enip

Zeek network security monitor plugin that enables parsing of the Ethernet/IP and Common Industrial Protocol standards
Zeek
45
star
48

amazon-pay-sdk-samples

Amazon Pay SDK Sample Code
PHP
43
star
49

oss-contribution-tracker

Track contributions made to external projects and manage CLAs
TypeScript
40
star
50

amazon-s3-gst-plugin

A collection of Amazon S3 GStreamer elements.
C
40
star
51

fashion-attribute-disentanglement

Python
39
star
52

zeek-plugin-s7comm

Zeek network security monitor plugin that enables parsing of the S7 protocol
Zeek
39
star
53

milan

Milan is a Scala API and runtime infrastructure for building data-oriented systems, built on top of Apache Flink.
Scala
39
star
54

selling-partner-api-samples

Sample code for Amazon Selling Partner API use cases
Python
37
star
55

orthogonal-additive-gaussian-processes

Light-weighted code for Orthogonal Additive Gaussian Processes
Python
37
star
56

jekyll-doc-project

This repository contains an open-source Jekyll theme for authoring and publishing technical documentation. This theme is used by Appstore/Alexa tech writers and other community members. Most of the theme's files are stored in a Ruby Gem (called jekyll-doc-project).
HTML
36
star
57

smoke-dynamodb

SmokeDynamoDB is a library to make it easy to use DynamoDB from Swift-based applications, with a particular focus on usage with polymorphic database tables (tables that do not have a single schema for all rows).
Swift
34
star
58

amazon-pay-api-sdk-nodejs

Amazon Pay API SDK (Node.js)
JavaScript
34
star
59

zeek-plugin-bacnet

Zeek network security monitor plugin that enables parsing of the BACnet standard building controls protocol
Zeek
30
star
60

ss-aga-kgc

Python
30
star
61

chalet-charging-location-for-electric-trucks

Optimization tool to identify charging locations for electric trucks
Python
30
star
62

amazon-pay-api-sdk-java

Amazon Pay API SDK (Java)
Java
29
star
63

credence-to-causal-estimation

A framework for generating complex and realistic datasets for use in evaluating causal inference methods.
Python
29
star
64

sparse-vqvae

Experimental implementation for a sparse-dictionary based version of the VQ-VAE2 paper
Python
28
star
65

zeek-plugin-profinet

Zeek network security monitor plugin that enables parsing of the Profinet protocol
Zeek
28
star
66

ion-tests

Test vectors for testing compliant Ion implementations.
25
star
67

differential-privacy-bayesian-optimization

This repo contains the underlying code for all the experiments from the paper: "Automatic Discovery of Privacy-Utility Pareto Fronts"
Python
25
star
68

buy-with-prime-cdk-constructs

This package extends common CDK constructs with opinionated defaults to help create an organization strategy around infrastructure as code.
TypeScript
25
star
69

basis-point-sets

Python
24
star
70

ion-hive-serde

A Apache Hive SerDe (short for serializer/deserializer) for the Ion file format.
Java
24
star
71

zeek-plugin-tds

Zeek network security monitor plugin that enables parsing of the Tabular Data Stream (TDS) protocol
Zeek
24
star
72

ion-intellij-plugin

Support for Ion in Intellij IDEA.
Kotlin
23
star
73

ion-schema-kotlin

A Kotlin reference implementation of the Ion Schema Specification.
Kotlin
23
star
74

smoke-framework-application-generate

Code generator to generate SmokeFramework-based applications from service models.
Swift
23
star
75

emukit-playground

A web page explaining concepts of statistical emulation and making decisions under uncertainty in an interactive way.
JavaScript
22
star
76

ftv-livetv-sample-tv-app

Java
22
star
77

smoke-framework-examples

Sample applications showing the usage of the SmokeFramework and related libraries.
Swift
22
star
78

ion-hash-go

A Go implementation of Amazon Ion Hash.
Go
22
star
79

pretraining-or-self-training

Codebase for the paper "Rethinking Semi-supervised Learning with Language Models"
Python
22
star
80

tiny-attribution-generator

A small tool and library to create attribution notices from various formats
TypeScript
20
star
81

confident-sinkhorn-allocation

Pseudo-labeling for tabular data
Jupyter Notebook
20
star
82

ion-docs

Source for the GitHub Pages for Ion.
Java
19
star
83

autotrail

AutoTrail is a highly modular, partial automation workflow engine providing run time execution control
Python
19
star
84

git-commit-template

Set commit templates for git
JavaScript
19
star
85

smoke-aws-generate

Code generator to generate the SmokeAWS library from service models.
Swift
18
star
86

amazon-codeguru-profiler-for-spark

A Spark plugin for CPU and memory profiling
Java
17
star
87

smoke-aws-credentials

A library to obtain and assume automatically rotating AWS IAM roles written in the Swift programming language.
Swift
17
star
88

service-model-swift-code-generate

Modular code generator to generate Swift applications from service models.
Swift
17
star
89

amazon-pay-api-sdk-dotnet

Amazon Pay API SDK (.NET)
C#
17
star
90

sample-fire-tv-app-video-skill

This sample Fire TV app shows how to integrate an Alexa video skill in a simple, basic way.
Java
16
star
91

amazon-template-library

A collection of general purpose C++ utilities that play well with the Standard Library and Boost.
C++
16
star
92

ion-cli

Rust
15
star
93

refuel-open-domain-qa

Python
15
star
94

rheoceros

Cloud-based AI / ML workflow and data application development framework
Python
15
star
95

amazon-instant-access-sdk-php

PHP SDK to aid in 3p integration with Instant Access
PHP
14
star
96

amazon-mcf-plugin-for-magento-1

Plugin code to enable Amazon MCF in Magento 1.
PHP
14
star
97

login-with-amazon-wordpress

A pre-integrated plugin that can be installed into a Wordpress powered website to integrate with Login with Amazon.
PHP
14
star
98

amzn-ec2-ena-utilities

Python
14
star
99

firetv-sample-touch-app

This sample Android project demonstrates how to build the main UI of a Fire TV application in order to support both Touch interactions and Remote D-Pad controls.
Java
13
star
100

eslint-plugin-no-date-parsing

Disallow string parsing with new Date and Date.parse.
TypeScript
13
star