Swiftpack.co - Package - watson-developer-cloud/swift-sdk

Watson Developer Cloud Swift SDK

Build Status Carthage Compatible Documentation CLA assistant wdc-community.slack.com


The Watson Developer Cloud Swift SDK makes it easy for mobile developers to build Watson-powered applications. With the Swift SDK you can leverage the power of Watson's advanced artificial intelligence, machine learning, and deep learning techniques to understand unstructured data and engage with mobile users in new ways.

There are many resources to help you build your first cognitive application with the Swift SDK:




This SDK provides classes and methods to access the following Watson services.

Before you begin


  • Xcode 9.3+
  • Swift 4.2+
  • iOS 10.0+

Migrating to version 3.0.0

Version 3.0.0 introduced breaking changes that affect all frameworks of the Swift SDK. Please refer to the migration guide for information on how to get up and running with the new version.


The IBM Watson Swift SDK can be installed with Cocoapods, Carthage, or Swift Package Manager.


You can install Cocoapods with RubyGems:

$ sudo gem install cocoapods

If your project does not yet have a Podfile, use the pod init command in the root directory of your project. To install the Swift SDK using Cocoapods, add the services you will be using to your Podfile as demonstrated below (substituting MyApp with the name of your app). The example below shows all of the currently available services; your Podfile should only include the services that your app will use.


target 'MyApp' do
    pod 'IBMWatsonAssistantV1', '~> 3.2.0'
    pod 'IBMWatsonAssistantV2', '~> 3.2.0'
    pod 'IBMWatsonCompareComplyV1', '~> 3.2.0'
    pod 'IBMWatsonDiscoveryV1', '~> 3.2.0'
    pod 'IBMWatsonLanguageTranslatorV3', '~> 3.2.0'
    pod 'IBMWatsonNaturalLanguageClassifierV1', '~> 3.2.0'
    pod 'IBMWatsonNaturalLanguageUnderstandingV1', '~> 3.2.0'
    pod 'IBMWatsonPersonalityInsightsV3', '~> 3.2.0'
    pod 'IBMWatsonSpeechToTextV1', '~> 3.2.0'
    pod 'IBMWatsonTextToSpeechV1', '~> 3.2.0'
    pod 'IBMWatsonToneAnalyzerV3', '~> 3.2.0'
    pod 'IBMWatsonVisualRecognitionV3', '~> 3.2.0'

Run the pod install command, and open the generated .xcworkspace file. To update to newer releases, use pod update.

When importing the frameworks in source files, exclude the IBMWatson prefix and the version suffix. For example, after installing IBMWatsonAssistantV1, import it in your source files as import Assistant.

For more information on using Cocoapods, refer to the Cocoapods Guides.


You can install Carthage with Homebrew:

$ brew update
$ brew install carthage

If your project does not have a Cartfile yet, use the touch Cartfile command in the root directory of your project. To install the IBM Watson Swift SDK using Carthage, add the following to your Cartfile.

github "watson-developer-cloud/swift-sdk" ~> 3.2.0

Then run the following command to build the dependencies and frameworks:

$ carthage bootstrap --platform iOS

Follow the remaining Carthage installation instructions here. Note that the above command will download and build all of the services in the IBM Watson Swift SDK. Make sure to drag-and-drop the built frameworks (only for the services your app requires) into your Xcode project and import them in the source files that require them. The following frameworks need to be added to your app:

  1. RestKit.framework
  2. Whichever services your app will be using (AssistantV1.framework, DiscoveryV1.framework, etc.)
  3. (Speech to Text only) Starscream.framework

If your app fails to build because it is built with a different version of Swift than the downloaded SDK, then re-run the carthage update command with the --no-use-binaries flag added.

Swift Package Manager

Add the following to your Package.swift file to identify the IBM Watson Swift SDK as a dependency. The package manager will clone the Swift SDK when you build your project with swift build.

dependencies: [
    .package(url: "https://github.com/watson-developer-cloud/swift-sdk", from: "3.2.0")


The Identity and Access Management (IAM) service of the IBM Cloud is the primary method of authentication to IBM Cloud services. Some service instances may use an alternate form of authentication, such as basic authentication (username and password).

Getting credentials

To find out which authentication to use, view the service credentials. You find the service credentials for authentication the same way for all Watson services:

  1. Go to the IBM Cloud Dashboard page.
  2. Either click an existing Watson service instance in your resource list or click Create resource > AI and create a service instance.
  3. Click on the Manage item in the left nav bar of your service instance.

On this page, you will see your credentials to use in the SDK to access your service instance.

Supplying credentials

The SDK provides separate init methods for each form of authentication that may be used by instances of the service.

  • For service instances that use IAM authentication, the SDK provides two init methods -- one that accepts an apikey and another that accepts an access token created from an apikey. If you use the init method that supplies the apikey, the SDK will obtain an access token and refresh it when needed. If you initialize the SDK with the method that supplies an access token, you will need to periodically refresh the token as they expire after a short time. Learn more about IAM.

  • For service instances that use basic authentication (username and password), use the init method that specifies the username and password.

Credentials in the environment or a local credentials file

The SDK can extract service credentails from the environment, e.g. the VCAP_SERVICES environment variable, or a local credentials file.

To use credentials stored in a local file, go to the Manage tab of your service instance on IBM Cloud, and click on the button to download the credentials. The file will be called ibm-credentials.env. Add this file to a location that is accessible from your project. For iOS apps, make sure to add it to the application target.

let discovery = Discovery(version: "your-version")

If your project is using multiple Watson services, you can merge the contents of the ibm-credentials.env files into a single file. Lines in the file can be added, deleted, or reordered, but the content of each line must not be changed.

Copy-Pasting Credentials

Copy the credentials from IBM Cloud and store them within your project. Then pass those values to the service initializer that accepts the type of credentials you have.


Some services use token-based Identity and Access Management (IAM) authentication. IAM authentication uses a service API key to get an access token that is passed with the call. Access tokens are valid for approximately one hour and must be regenerated.

You supply either an IAM service API key or an access token:

  • Use the API key to have the SDK manage the lifecycle of the access token. The SDK requests an access token, ensures that the access token is valid, and refreshes it if necessary.
  • Use the access token if you want to manage the lifecycle yourself. For details, see Authenticating with IAM tokens. If you want to switch to API key, override your stored IAM credentials with an IAM API key.
Supplying the IAM API key
let discovery = Discovery(version: "your-version", apiKey: "your-apikey")

If you are supplying an API key for IBM Cloud Private (ICP), use basic authentication instead, with "apikey" for the username and the api key (prefixed with icp-) for the password. See the Username and Password section.

Supplying the accessToken
let discovery = Discovery(version: "your-version", accessToken: "your-accessToken")
Updating the accessToken
Username and Password
let discovery = Discovery(version: "your-version", username: "your-username", password: "your-password")

Custom Service URLs

You can set a custom service URL by modifying the serviceURL property. A custom service URL may be required when running an instance in a particular region or connecting through a proxy.

For example, here is how to connect to a Tone Analyzer instance that is hosted in Germany:

let toneAnalyzer = ToneAnalyzer(
    version: "yyyy-mm-dd",
    username: "your-username",
    password: "your-password"
toneAnalyzer.serviceURL = "https://gateway-fra.watsonplatform.net/tone-analyzer/api"

Disable SSL certificate verification

For ICP(IBM Cloud Private), you can disable the SSL certificate verification by:


Note: disableSSLVerification() is currently not supported on Linux.

Custom Headers

There are different headers that can be sent to the Watson services. For example, Watson services log requests and their results for the purpose of improving the services, but you can include the X-Watson-Learning-Opt-Out header to opt out of this.

We have exposed a defaultHeaders public property in each class to allow users to easily customize their headers:

let naturalLanguageClassifier = NaturalLanguageClassifier(username: username, password: password)
naturalLanguageClassifier.defaultHeaders = ["X-Watson-Learning-Opt-Out": "true"]

Each service method also accepts an optional headers parameter which is a dictionary of request headers to be sent with the request.

Featured Projects

We love to highlight cool open-source projects that use this SDK! If you'd like to get your project added to the list, feel free to make an issue linking us to it.

Synchronous Execution

By default, the SDK executes all networking operations asynchronously. If your application requires synchronous execution, you can use a DispatchGroup. For example:

let dispatchGroup = DispatchGroup()
assistant.message(workspaceID: workspaceID) { response, error in
	if let error = error {
    if let message = response?.result else {
dispatchGroup.wait(timeout: .distantFuture)

Handling PNG and CgBI Files in XCode

When working with iOS projects, you may need to add PNG images to your XCode resource bundle. By default, XCode converts PNG files to Apple's CgBI file format as an optimization step.

Watson services that accept PNG images as input files (Visual Recognition, Compare and Comply) are not able to process CgBI files. In some cases passing PNGs that have been encoded as CgBI will result in HTTP errors, and in other cases (Visual Recognition) will result in a 200 Success with no results.

How to bypass CgBI Encoding for Resource Files

In order to bypass CgBI encoding and keep the PNG files in a format that will operate well with Watson services, select the PNG file in XCode, and modify the Type attribute to Data in the File Inspector.

File inspector example

Future plans for handling CgBI

In the future, we will explore our options to handle conversion between CgBI and PNG within the Swift SDK, but this is currently not available.

Objective-C Compatibility

Please see this tutorial for more information about consuming the Watson Developer Cloud Swift SDK in an Objective-C application.

Linux Compatibility

To use the Watson SDK in your Linux project, please follow the Swift Package Manager instructions.. Note that Speech to Text and Text to Speech are not supported because they rely on frameworks that are unavailable on Linux.


We would love any and all help! If you would like to contribute, please read our CONTRIBUTING documentation with information on getting started.


This library is licensed under Apache 2.0. Full license text is available in LICENSE.

This SDK is intended for use with an Apple iOS product and intended to be used in conjunction with officially licensed Apple development tools.


Stars: 872


Used By

Total: 0


v3.2.0 - 2020-01-20 21:32:15

3.2.0 (2020-01-20)

Bug Fixes

  • hand edit: keep model in CategoriesOptions (b0ceeb5)
  • hand edit: parse "collectionIDs" and "features" into csv data (4e32293)


  • environment auth: make environment authenticator a throwable (35c8d30)
  • generated code: update based on api defs (e779c04)
  • hand edit: add discriminator support for QueryAggregation (e3f065d)
  • hand edit: make discoveryv2 environment init throwable (1510586)
  • hand edit: make VR V4 environment init throwable (4e63c19)
  • websockets: add new function params (80424fd)

3.1.0 - 2019-11-27 20:57:22

3.0.0 - 2019-10-04 19:21:58

3.0.0 (2019-10-04)

Bug Fixes

  • hand edit: parse "collectionIDs" and "features" into csv data (7054591)
  • hand edit: vR V4 model hand edits (27d17b2)
  • linux tests: fix syntax errors in linux tests (bfb2b62)

Build System

  • core: update dependencies to use new core (dbfdb19)

Code Refactoring

  • remove utility methods: remove authentication type methods (81ab28e)


  • environment auth: make environment authenticator a throwable (8689481)
  • environment auth: make environment authenticator a throwable (e9df576)
  • generated code: add generated code based on updated api definitions (e9a8eeb)
  • generated code: update based on api reference changes (de24497)
  • generated code: update generated code based on api definitions (e090f9d)
  • hand edit: make VR V4 environment init throwable (52788bc)
  • vis rec v4: add Visual Recognition V4 (96e6dab)
  • websockets: accept authenticator in websocket client init (db90f26)


  • core: uses new core framework
  • generated code: new authentication scheme
  • environment auth: Previously, this was an init? but now it will be an init -> throws
  • websocketURL is now a String?

refactor(core import): use IBMSwiftSDKCore instead of RestKit

feat(vis rec ui): remove faces method

  • faces method is no longer available

build(xcode): update dependency settings

test(vr): update test values

fix(hand edit): convert classifierIDs and owners to csv data

  • environment auth: Previously, this was an init? but now it will be an init -> throws
  • hand edit: turns the init? into an init -> throws
  • faces method is no longer available

build(xcode): update dependency settings

  • generated code: removed models and breaking api changes
  • websockets: new init parameters and removal of convience inits
  • remove utility methods: public methods removed
  • websocketURL is now a String?

refactor(core import): use IBMSwiftSDKCore instead of RestKit

feat(vis rec ui): remove faces method

2.3.0 - 2019-08-27 16:49:11

2.3.0 (2019-08-27)


  • comparecomply: update based on latest api definition (a8be492)

2.2.0 - 2019-07-31 17:22:06

2.2.0 (2019-07-31)


  • AssistantV2 add support for Assistant Search Skill (ca7e3b0)

Bug fixes

  • Integration Tests various bug fixes

2.1.1 - 2019-06-28 20:42:33

2.1.1 (2019-06-28)

Bug Fixes

  • SpeechToTextV1: Add configureSession parameter to RecognizeMicrophone. (87f5aed)

2.1.0 - 2019-06-14 05:38:39

2.1.0 (2019-06-14)


  • Add support for insecure connections (975bb68)
  • Generated code for June SDK release (7283310)
    • Assistant V1 updated DialogNodeOutputOptionsElementValue
    • LTv3 Document Translation API
    • NLU categories explanations
    • STT processing metrics and audio metrics features
  • SpeechToTextV1: Add support for processing metrics and audo metrics to websockets methods (e3b7169)

Bug Fixes

  • Remove support for insecure connections for Linux (76f15f2)

2.0.3 - 2019-05-13 20:34:41

2.0.3 (2019-05-13)

Bug Fixes

  • Use full url in semantic release (2c637be)

2.0.2 - 2019-04-16 01:34:47

Bug Fixes

  • core: Look for iam_apikey in credential file with IAM auth (689d748 (https://github.com/watson-developer-cloud/swift-sdk/commit/689d748))
  • Fix swiftlint errors (9b12dba (https://github.com/watson-developer-cloud/swift-sdk/commit/9b12dba))

2.0.1 - 2019-04-08 11:20:09

Bug Fixes

  • VisualRecognitionV3:
    • Add .zip to filenames passed to create/update classifier
    • Fix multiple calls to completionHandler from classifyWithLocalModel.

2.0.0 - 2019-03-29 03:07:04


  • All:
    • Version moved to first parameter of init methods for basic authentication
    • Credentials file support in the intializer dropped for iOS. Only supported in Linux
    • Error response handling reworked for consistency across services and languages
    • Ordering of parameters in some methods has changed due to migration of the API docs and generator to OpenAPI 3.0
    • File-type parameters are now defined as Data rather than URL
  • AssistantV1:
    • Some model classes have been merged / simplified
    • The InputData class has been renamed to MessageInput
  • SpeechToTextV1:
    • The recognizeWithWebsockets method has been revised to accept a RecognizeCallback object with onResults and onError callback properties

1.4.0 - 2019-02-12 18:56:33

1.4.0 (2019-02-12)

Bug Fixes

  • AssistantV1: Remove erroneous additionalProperties from MessageResponse (60f1616)
  • CompareComplyV1: BodyCells had incorrect types for some of its properties (13c66af)


  • All: New initializer that loads credentials from file (39ddae3)
  • AssistantV1: Add additionalProperties property to InputData (1e14888)
  • CompareComplyV1: Add address as a possible Attribute (b34a782)
  • CompareComplyV1: Add attributes property to BodyCells (3302504)
  • CompareComplyV1: Add importance to Parties (96eb790)
  • CompareComplyV1: Add confidence level for the identification of the contract amount (35a773a)
  • DiscoveryV1: Add method to get the stopword list status (4a6b615)
  • DiscoveryV1: Add the pending status for Documents (f342c68)
  • SpeechToTextV1: Add new parameter force to upgradeAcousticModel() method (eecf18b)

1.3.1 - 2019-01-18 17:52:36

1.3.1 (2019-01-18)

Bug Fixes

  • SpeechToTextV1: Fix grammarName and redaction parameters in recognize websocket methods (64b116c)

1.3.0 - 2019-01-18 01:37:31

1.3.0 (2019-01-18)

Bug Fixes

  • SpeechToTextV1: Change contentType parameter to optional in certain methods (e033cff)


  • DiscoveryV1: Add support for custom stopword lists (915ce68)
  • DiscoveryV1: Add support for gateways (39393fa)
  • DiscoveryV1: Add web crawlers to the list of possible sources (5a4a62e)
  • SpeechToTextV1: Add new options to acoustic models and language models (3345b46)
  • SpeechToTextV1: Add the ability to specify grammars in recognition requests (7edcdf4)
  • VisualRecognitionV3: Add acceptLanguage parameter to detectFaces() (a260a9c)
  • VisualRecognitionV3: Add genderLabel property to FaceGender model (a00f3c6)

1.2.0 - 2019-01-11 04:19:13

1.2.0 (2019-01-11)

Bug Fixes

  • CompareComplyV1: Change Location properties to optional (2e66ac5)
  • CompareComplyV1: Fix incorrect parameter types (4cfa292)
  • CompareComplyV1: Give more appropriate types to model properties (4b1af08)


  • CompareComplyV1: Add properties to AlignedElements and Attribute (0fbeb6d)
  • CompareComplyV1: New framework for Compare & Comply service (482444a)

1.1.1 - 2019-01-10 00:48:21

1.1.0 (2019-01-10)

Bug Fixes

  • AssistantV1: Add missing "disabled" field to DialogNode (e45de83)
  • AssistantV2: Add missing userDefined field to MessageOutput (f65cafc)

1.1.0 - 2018-12-11 21:26:40

1.1.0 (2018-12-11)


  • AssistantV1: Add metadata field to Context model (13a90c1)
  • AssistantV1: Add option to sort results in getWorkspace() (5cefc7b)
  • DiscoveryV1: Add new concepts property to NluEnrichmentFeatures model (80258db)
  • DiscoveryV1: Add retrievalDetails property to QueryResponse model (631affc)
  • NaturalLanguageUnderstandingV1: Add 4 new properties to the Model model (53fe057)
  • NaturalLanguageUnderstandingV1: Add new count property to KeywordsResult model (ab9a339)
  • NaturalLanguageUnderstandingV1: Add new limit property to CategoriesOptions model (5bf6637)

1.0.0 - 2018-12-06 21:51:37

1.0.0 (2018-12-06)

All Services

  • failure and success callbacks are replaced with a single completionHandler of type (WatsonResponse<T>?, WatsonError?) -> Void
  • New WatsonResponse type in the completion handlers that contains the response HTTP status, headers, and data
  • New WatsonError type in the completion handlers that contains more useful and detailed information about the error that occurred
  • Change the type of date-time properties from String to Date
  • Remove all deprecated types, methods, and properties
  • All parameters now get passed directly to methods rather than packaging them up in the properties parameter
  • All models are now Codable, instead of only being Encodable or Decodable, so they can be converted both to and from JSON
  • All models are now Equatable


  • REMOVED - use AssistantV1 instead


  • REMOVED - use LanguageTranslatorV3 instead


  • Remove additionalProperties property from CategoriesOptions and MetadataOptions


  • The tokenURL and websocketsURL properties are no longer public. Setting the serviceURL will automatically update the other two URLs appropriately.
  • Add missing parameters baseModelVersion, languageCustomizationID, and customerID to the following methods:
    • recognize() (using an audio file)
    • recognizeUsingWebsocket()
    • recognizeMicrophone()
    • SpeechToTextSession() initializers


  • Remove PositiveExample model. All parameters that used that model are now of type [String: URL]
  • Change the following properties from optional to nonoptional
    • ClassResult.score
    • DetectedFaces.imagesProcessed
    • FaceAge.score
    • FaceGender.score

0.38.1 - 2018-11-13 23:29:47

0.38.1 (2018-11-13)

Bug Fixes

  • SpeechToTextV1: Update recognizeMicrophone() to work with any authentication method (5701ba6)

0.38.0 - 2018-11-12 22:52:45

0.38.0 (2018-11-12)

Bug Fixes

  • VisualRecognitionV3: temporary workaround for new A12 based devices (62edd09)
  • VisualRecognitionV3: temporary workaround to support A12 devices for Core ML inference (fa212ec)


  • All: Add support for Swift 4.2 (4bbf42b)

0.37.0 - 2018-11-02 15:52:35

0.37.0 (2018-11-02)


  • AssistantV1, AssistantV2: Add cloudFunction and webAction to DialogNodeAction.ActionType (16d3fc9)
  • DiscoveryV1: Tokenization dictionaries for collections (d274371)
  • SpeechToTextV1: Add languageCustomizationID parameter to createJob() and recognize() (0137964)

0.36.0 - 2018-10-19 01:38:51

0.36.0 (2018-10-19)

Bug Fixes

  • Visual Recognition: Fix deserialization error in getCoreMlModel (9392b23)
  • Conversion of file data to multipart form data (1d46baf)


  • Discovery: Add "LT" option to environment sizes (f92fcde)
  • Discovery: Add size parameter to updateEnvironment method (725e1d5)
  • Discovery: Add bias and loggingOptOut parameters to query methods (8782fc6)
  • Discovery: Add requestedSize and searchStatus properties to Environment model (14cdb02)
  • ToneAnalyzer: Add ToneID options to ToneChatScore (ac75c92)


  • AssistantV1: Revert erroneous addition of actions property to OutputData (28efe1e)
  • Linux: Remove 30 minute wait from Linux tests (1b4d734)

- 2018-09-25 19:58:56

Adds a new framework for the AssistantV2 service

- 2018-09-14 15:22:15

This release regenerates all services with documentation updates, adds support for IBM Cloud Private authentication, and adds a couple of new properties to Speech to Text and Assistant/Conversation.

  • All services: Adds support for authentication with IBM Cloud Private (ICP) API keys. Use the init(username:password:) initializer, passing apikey for the username parameter and the API key for the password.
  • Speech to Text: Adds the language models de-DE_BroadbandModel, pt-BR_BroadbandModel, and pt-BR_NarrowbandModel
  • Assistant/Conversation: Add the actions property to MessageResponse and OutputData

- 2018-09-11 22:15:45

Internal bug fix regarding RestKit.framework. This should have no impact on users.

- 2018-08-31 19:14:43

First release that supports Cocoapods!

This release adds support for installing each service in the Swift SDK via Cocoapods. It also adds new options to the Speech to Text recognize API, and includes documentation improvements.

  • Support for Cocoapods!
  • Speech to Text: Adds the acousticCustomizationID and headers parameters to the recognize() method
  • Speech to Text: Adds the customizationWeight property to RecognitionSettings
  • Documentation: Split README into multiple READMEs: one main README plus one README for each service
  • Documentation: Improved Jazzy docs

v0.32.0 - 2018-08-17 14:57:36

This release regenerates all services with documentation updates, and adds some new methods to Discovery.

  • Discovery: Add ability to create new Events to create log entries associated with specific queries
  • Discovery: Add new methods to obtain metrics and metadata on past queries
  • Discovery: Changes the type of the size parameter in the createEnvironment method from Int to String.

v0.31.0 - 2018-08-08 18:05:04

This release regenerates all services with documentation updates and minor generator improvements.

This release includes the following new features and bug fixes:

  • Assistant: Add support for entity mentions and disambiguation
  • Renames STT Websocket method to recognizeUsingWebSocket. The prior name is deprecated but retained for backward compatibility.
  • Fix Xcode 10 build error

v0.30.0 - 2018-07-20 19:42:29

This release regenerates all services with documentation updates and minor generator improvements.

This release includes the following new features and bug fixes:

  • Discovery: Add support for credentials and configuration source options
  • Add support for websocket features for services using IAM authentication
  • Handle malformed URLs when constructing URL requests
  • Remove extraneous version param from IAM init methods for LTv2, NLC, STT, TTS

v0.29.0 - 2018-06-29 15:06:34

This release regenerates all services with generator updates for error handling and multi-consumes operations.

The release also adds a deprecation notice for the Language Translator v2 service. Users are encouraged to migrate to the new Language Translator v3 service.

This release should be fully compatible with the previous SDK release.