[FFL-3355] Add standalone rules configuration fetch helper

packages/react-native-openfeature/README.md

@@ -338,6 +338,50 @@ precomputed-only `configurationFromString` helper.
   `Error`. Loading a usable configuration after an error emits `Ready`, followed by
   `ConfigurationChanged`.
 
+### Fetching rules configuration
+
+Use `fetchRulesConfiguration` from the `/rules-based` entry point to fetch client-distributed rules and decode them into a `FlagsConfiguration`. Like the browser helper, it accepts `clientToken`, `env`, an optional site domain, an optional `flagSelection`, an optional `fetch` implementation, and an optional `AbortSignal`. An optional `logger` receives selection diagnostics and defaults to `console`. The helper does not require native SDK initialization or `DdFlags.enable()`.
+
+```tsx
+import {
+    DatadogCoreProvider,
+    fetchRulesConfiguration
+} from '@datadog/mobile-react-native-openfeature/rules-based';
+import { OpenFeature } from '@openfeature/react-sdk';
+
+const fetchOptions = {
+    clientToken: '<CLIENT_TOKEN>',
+    env: 'production',
+    site: 'datadoghq.com',
+    flagSelection: { tags: ['team:checkout'], keyPatterns: ['checkout-*'] }
+};
+const provider = new DatadogCoreProvider();
+provider.setConfiguration(await fetchRulesConfiguration(fetchOptions));
+await OpenFeature.setProviderAndWait('datadog-core', provider, {
+    targetingKey: 'user-123', country: 'US'
+});
+const client = OpenFeature.getClient('datadog-core');
+const enabled = client.getBooleanValue('checkout-new-flow', false);
+
+// Context changes evaluate the loaded rules locally, without another request.
+await OpenFeature.setContext('datadog-core', {
+    targetingKey: 'user-456', country: 'CA'
+});
+
+// Refresh explicitly when the application needs new rules.
+provider.setConfiguration(await fetchRulesConfiguration(fetchOptions));
+```
+
+The helper makes one request per call, with no automatic polling, retries, persistence, or tracking. It rejects on HTTP, network, cancellation, body-read, or decoding failures. Catch refresh errors before replacing the provider configuration to keep evaluating the last successfully loaded rules. Use a client token; the helper only requests client-distributed configuration, never server-distributed rules.
+
+`flagSelection` requests flags matching any of the listed tags or key patterns (OR). Tags are exact, case-sensitive matches; key patterns are case-sensitive and match the whole key, with `*` matching any run of characters. Values are trimmed using Unicode whitespace, deduplicated, and sorted by UTF-8 bytes. Each list accepts at most 31 distinct nonempty values after normalization, and each value can contain at most 200 UTF-8 bytes. Empty selections are omitted quietly.
+
+Selection is best-effort: a malformed or over-limit selection, or one that makes the encoded request URL exceed 8,192 bytes, is not sent. The helper logs an error and requests the full configuration instead. A nonempty `DD-Flag-Selection-Ignored` response header also produces an error diagnostic when a selection was sent. Supply `logger: { error(message) { /* route diagnostics */ } }` to customize reporting. Diagnostics do not initialize the native SDK or send telemetry, and a throwing logger does not prevent configuration loading.
+
+When the backend applies selection, flags outside it are absent from the returned configuration. Evaluating them returns your default value with `FLAG_NOT_FOUND`, so select every flag the app evaluates. Selection is not an access-control boundary: the helper retains the full returned configuration if the server ignores selection and does not filter it locally.
+
+`site` uses a domain such as `datadoghq.com` or `datadoghq.eu`, not the native SDK's `US1` or `EU1` identifiers. The default is `datadoghq.com`. Fetching and evaluation do not register tracking; use the hooks below when tracking is needed.
+
 ### Optional tracking with DatadogCoreProvider (Android only)
 
 Register exposure, evaluation, and RUM hooks independently with the evaluation-only

packages/react-native-openfeature/package.json

@@ -36,16 +36,17 @@
     "scripts": {
         "test": "jest",
         "lint": "eslint .",
-        "prepare": "rm -rf lib && yarn bob build"
+        "prepare": "genversion --es6 --semi src/version.ts && rm -rf lib && yarn bob build"
     },
     "dependencies": {
-        "@datadog/flagging-core": "3.1.1"
+        "@datadog/flagging-core": "3.3.0"
     },
     "devDependencies": {
         "@datadog/mobile-react-native": "^3.11.0",
         "@openfeature/core": "^1.9.2",
         "@openfeature/web-sdk": "^1.8.0",
         "@testing-library/react-native": "7.0.2",
+        "genversion": "3.0.2",
         "react-native-builder-bob": "0.26.0",
         "react-native-gesture-handler": "1.10.3"
     },

packages/react-native-openfeature/src/__tests__/__utils__/flag-selection/SOURCE.md

@@ -0,0 +1,5 @@
+# Shared Flag Selection Vectors
+
+`selection-vectors.json` is copied unchanged from [DataDog/ffe-system-test-data](https://github.com/DataDog/ffe-system-test-data/blob/caa6d1ac3e5fbda0ce57ef3ffff3772901f3f466/flag-selection/selection-vectors.json) at commit `caa6d1ac3e5fbda0ce57ef3ffff3772901f3f466`.
+
+The fetch-helper tests consume all 38 `sdk` expectations. Raw-query percent-decoding and server-side `selectedKeys` expectations are not SDK transport behavior and are not exercised here. Test runs use this checked-in snapshot without network access. Update shared expectations upstream, then replace this file from a pinned commit and update this provenance and the SDK-case count assertion together.