Add native FF&E configuration flow for React Native

[leoromanovsky]

Motivation

We need a React Native delivery path for offline initialization and dynamic configuration that can move quickly for a customer evaluation without first landing, reviewing, releasing, and consuming upstream iOS and Android SDK changes. This branch keeps the implementation native-first: React Native exposes the RFC-shaped rules/offline configuration API surface, while Kotlin and Swift own configuration parsing, evaluator state, rules fetch, persistence, and SDK side effects in RN-local code that can later be extracted into the iOS and Android SDKs.

Changes

• Adds the native rules/offline configuration API surface in React Native: parse/serialize config wire, fetch rules config, save/load, explicit activation, dynamic context, typed evaluations, and debug state.
• Implements RN-local Kotlin and Swift UFC evaluators that parse configuration once into native models and evaluate from native state, avoiding JSON deserialization on every flag evaluation.
• Reuses existing native Flags SDK side-effect hooks for exposures, flag-evaluation EVP emission, and RUM feature-flag annotation instead of adding a JS side-effect pipeline.
• Adds shared ffe-system-test-data parity tests plus a new-architecture Android smoke that runs the offline corpus from React Native JS through the native bridge in CI.

Decisions

• Phase 1 is RN-local but native-first so React Native can move now, while the Kotlin/Swift pieces remain shaped for later extraction into the iOS and Android SDKs.
• The existing DdFlags / FlagsClient / DatadogOpenFeatureProvider precompute client remains unchanged; this PR adds a separate rules/offline path.
• Fetch and load are intentionally non-mutating. Active evaluation state changes only through explicit setConfiguration.
• Shared fixtures are the cross-repo correctness contract for this branch and the future iOS/Android extraction.

React Native API Example

import{CoreConfiguration,DdSdkReactNative,TrackingConsent,}from'@datadog/mobile-react-native';constsdkConfig=newCoreConfiguration('<client-token>','staging',TrackingConsent.GRANTED,{rumConfiguration: {applicationId: '<rum-application-id>',},},);awaitDdSdkReactNative.initialize(sdkConfig);// Offline initialization from a bundled rules ConfigurationWire string.constofflineConfig=awaitDdSdkReactNative.configurationFromString(bundledRulesConfigurationWire,);awaitDdSdkReactNative.setConfiguration(offlineConfig);awaitDdSdkReactNative.setEvaluationContext({targetingKey: user.id,attributes: {plan: user.plan,},});constcheckoutEnabled=awaitDdSdkReactNative.resolveBooleanEvaluation('checkout.enabled',false,);// Later, refresh the rules config without changing active evaluation state.constpreviousWire=awaitDdSdkReactNative.configurationToString(offlineConfig);constfetchedConfig=awaitDdSdkReactNative.fetchRulesConfiguration({endpoint: 'https://app.datadoghq.com/api/v2/feature-flagging/config/rules-based',headers: {'dd-client-token': '<client-token>',},flagQueryParams: {dd_env: 'staging',},previousConfigurationWire: previousWire,});awaitDdSdkReactNative.saveConfiguration(fetchedConfig,{slot: 'default'});// Activate the refreshed config only when the app is ready to switch.awaitDdSdkReactNative.setConfiguration(fetchedConfig);// Evaluations after activation use the refreshed rules config and current context.constrefreshedCheckoutEnabled=awaitDdSdkReactNative.resolveBooleanEvaluation('checkout.enabled',false,);

Flow

flowchart TB
RN[React Native app] --> API[New native rules/offline config APIs]
API --> Kotlin[Kotlin RN-local FF&E core]
API --> Swift[Swift RN-local FF&E core]
Kotlin --> ParsedK[Parsed UFC model]
Swift --> ParsedS[Parsed UFC model]
Kotlin --> NativeSDK[Existing native Flags SDK hooks]
Swift --> NativeSDK
NativeSDK --> EVP[EVP exposures and flag evaluations]
NativeSDK --> RUM[RUM feature-flag annotations]
Kotlin -. later extract .-> AndroidSDK[Android SDK library]
Swift -. later extract .-> IosSDK[iOS SDK library]

Loading

sequenceDiagram
participant JS as React Native JS
participant Bridge as Native bridge
participant Core as Native FF&E core
participant HTTP as Rules config API
participant Disk as Native storage
JS->>Bridge: configurationFromString(bundled wire)
Bridge->>Core: parse UFC once
JS->>Bridge: setConfiguration(config)
JS->>Bridge: setEvaluationContext(context A)
JS->>Bridge: resolve*Evaluation(flag, default)
Bridge-->>JS: value, reason, metadata
JS->>Bridge: setEvaluationContext(context B)
JS->>Bridge: resolve*Evaluation(flag, default)
JS->>Bridge: fetchRulesConfiguration(options)
Bridge->>HTTP: GET rules-based config
HTTP-->>Bridge: ConfigurationWire
Bridge-->>JS: fetched config, active state unchanged
JS->>Bridge: saveConfiguration(fetched config)
Bridge->>Disk: persist wire
JS->>Bridge: setConfiguration(fetched config)
JS->>Bridge: resolve*Evaluation(flag, default)

Loading

flowchart LR
CI[test:native-android] --> Metro[Metro]
CI --> Emulator[Android emulator]
Metro --> App[New architecture example app]
Emulator --> App
App --> Runner[offline fixture runner]
Runner --> Bridge[RN bridge]
Bridge --> Core[Kotlin native FFE core]
Core --> Result[233 shared fixture cases pass]

Loading

Validation

• yarn tsc -p example-new-architecture/tsconfig.json --noEmit
• yarn run lint passes with 0 errors and the existing 10 warnings.
• ANDROID_HOME=/opt/homebrew/share/android-commandlinetools ANDROID_SDK_ROOT=/opt/homebrew/share/android-commandlinetools JAVA_HOME=/Users/leo.romanovsky/.sdkman/candidates/java/17.0.13-tem ./gradlew --no-daemon clean build -PDdSdkReactNative_minSdkVersion=24 from packages/core/android.
• xcodebuild -workspace example-new-architecture/ios/DdSdkReactNativeExample.xcworkspace -scheme DatadogSDKReactNative-Unit-Tests test -destination 'platform=iOS Simulator,id=DACA5ADA-1149-4987-9ED2-1F76A786B332' passes, 193 tests including the Swift shared-corpus evaluator test.
• ANDROID_HOME=/opt/homebrew/share/android-commandlinetools ANDROID_SDK_ROOT=/opt/homebrew/share/android-commandlinetools JAVA_HOME=/Users/leo.romanovsky/.sdkman/candidates/java/17.0.13-tem ./example-new-architecture/scripts/native-ffe-offline-android-smoke.sh passes and renders Native FFE offline fixture pass: 233 cases across 30 files.
• This same Android smoke is wired into test:native-android, so CI now exercises the offline corpus from React Native JS through the native Android bridge without credentials.

[datadog-official]

🎉 All green!

🧪 All tests passed
❄️ No new flaky tests detected

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: f423aa9 | Docs | Datadog PR Page | Give us feedback!}