[PerfectSlayer]

What Does This Do

Adds com.datadog.openfeature.Provider, its evaluator (DDEvaluator), hooks (metrics, logging, span enrichment) and FeatureFlagsRuntime wiring configuration delivery and event pipelines together. Also ports #12774:

• exposures are sent from a finally-stage hook once the evaluation succeeded, deduplicated by a provider-side cache that replaces the exposure pipeline cache;
• applications can record exposures with the public ExposureHook;
• UFC split features are added to the flag metadata;
• requires OpenFeature SDK 1.23.0, so provider hooks run inside a MultiProvider.

The SDK is still not published; the agent-coupled provider remains the shipped one.

Motivation

Completes the standalone SDK (APMLP-1512).

Additional Notes

Most of this PR moves code from master's products/feature-flagging/feature-flagging-api (datadog.trace.api.openfeature). Review carefully the new internal/FeatureFlagsRuntime, the new Provider / DDEvaluator wiring (agent bridge SPI and Settings instead of FeatureFlaggingGateway, bootstrap classes and reflection), and the lazy OTel resolution in FlagEvalMetrics.

New files

File	Notes
internal/FeatureFlagsRuntime	Shared runtime, reference-counted per class loader (acquire() / release()), replacing the agent-side gateway. Picks the configuration source from Settings (Remote Config through the bridge, failing without the agent, or the CDN) and the event transport (agent event proxy, direct intake, or proxy → direct fallback). Owns the exposure and flag evaluation pipelines (evaluation pipeline skipped by the evaluation counts kill switch, both skipped without transport). Fans configuration out to listeners and replays it to late ones. A failed start is rolled back and not cached
internal/FeatureFlagsRuntimeTest	Product disabled, Remote Config without agent, agentless without API key, kill switch, both pipelines through the proxy, listener replay / removal, reference counting, failed start, exposure gating
internal/evaluation/ConfigurationStore	Atomic configuration snapshot + ready latch used by DDEvaluator, previously fields of the old DDEvaluator
CanonicalFixtureTest	Cross-SDK canonical fixture cases extracted from master's DDEvaluatorTest. Reads the ffe-system-test-data submodule (dd-smoke-tests/openfeature/src/test/resources/ffe-system-test-data) with JsonReading instead of Moshi
FlagEvalMetricsLateSdkForkedTest	An evaluation before the OTel SDK registration does not claim GlobalOpenTelemetry, and metrics flow once the SDK is set

Moved, package rename only (no logic change)

• Evaluator, FlagEvalMetricsHook
• Tests: FlagEvalMetricsHookTest, FlagEvalMetricsTest

Moved with significant changes

New file	Master origin	Changes
Provider	Provider	Constructor calls AgentBridges.detect() and Settings.load(bridge), builds DDEvaluator directly (reflective loadEvaluatorClass() / EVALUATOR_IMPL removed). Span enrichment gate read from Settings (SpanEnrichmentGate removed); SpanEnrichmentHook gets bridge.spanEnricher(). FlagEvalLoggingHook always added, resolving its writer lazily from the runtime; LinkageError guard dropped. FatalError message now uses the cause message. Public constructors and Options unchanged
DDEvaluator	DDEvaluator	Pure evaluation now lives in internal/evaluation/FlagEvaluator (#12840). Kept: evaluate() delegating to FlagEvaluator (OpenFeature types end to end, no conversion), exposure building (via the FlagEvaluator ExposureListener) and recording, context snapshot / flatten / prune helpers. Lifecycle: package-private (Runnable, AgentBridge, Settings) constructor; synchronized initialize() / shutdown() acquire / release FeatureFlagsRuntime and register a Consumer<ServerConfiguration> (was FeatureFlaggingGateway.activate() / ConfigListener); ConfigurationStore replaces latch + AtomicReference. Removed: agent compatibility probes (SPLIT_SERIAL_ID_SUPPORTED, USE_LEGACY_EXPOSURE_API, legacy ExposureEvent constructor) and static span enrichment gate. Exposures go to runtime.recordExposure() instead of FeatureFlaggingGateway.dispatch()
FlagEvalMetrics	FlagEvalMetrics	Was an eager one-time GlobalOpenTelemetry.get(). Now resolves the counter lazily on record() with getOrNoop(), rebuilds it when the global instance changes, falls back to get() only for OTel API < 1.57 or a legacy OTel javaagent bridge, disables itself on LinkageError. Synchronized shutdown() with a closed flag
FlagEvalLoggingHook	FlagEvalLoggingHook	Single Supplier<FlagEvaluationWriter> constructor; INSTANCE singleton, gateway constructors and isFlagEvaluationEnqueueEnabled() check removed
SpanEnrichmentHook	SpanEnrichmentHook	Calls the bridge SpanEnricher (serialId / runtimeDefault) instead of dispatching SpanEnrichmentEvent on the gateway. Capture logic unchanged
jmh/FlagEvalHookHotPathBenchmark	same	Gateway enqueue toggle and @TearDown removed, writer injected through a supplier, run command now :dd-openfeature:jmh
ProviderTest	ProviderTest	Gateway / writer stubs replaced by fake bridges (ProxyAgentBridge, LocalHttpServer, TestSettings). Internal API linkage failure tests removed. New: product disabled, Remote Config without agent, end-to-end evaluation posted through the pipeline
DDEvaluatorTest	DDEvaluatorTest	Agent compatibility, legacy exposure, gateway activation and allocation date ABI tests removed; canonical fixtures moved to CanonicalFixtureTest. Exposures captured by overriding recordExposure. New: type compatibility, container cycles in context copy / flatten, empty and scalar values
DDEvaluatorSpanEnrichmentTest	DDEvaluatorSpanEnrichmentForkedTest	No longer forked: the gate comes from TestSettings instead of a static. Tests for agents without serial id support removed
FlagEvalLoggingHookTest	FlagEvalLoggingHookTest	Mockito writer; enqueue-disabled and gateway consent tests removed
SpanEnrichmentHookTest	SpanEnrichmentHookTest	Capturing SpanEnricher replaces the gateway listener

Shared files

• dd-openfeature/build.gradle.kts: adds api("io.opentelemetry:opentelemetry-api:1.66.0"); OTel API becomes a hard transitive dependency (master had it compileOnly 1.47.0 and handled its absence)
• dd-openfeature/gradle.lockfile: updated for the OTel dependency

Stack

The full change is in #12709. It was split into this stack to ease the review.

1. Resolve Feature Flags enablement from instrumenter config #12845 (merged)
2. Add dd-openfeature SDK flag configuration and evaluation core #12840 (merged)
3. Add dd-openfeature SDK settings and configuration delivery #12841
4. Add dd-openfeature SDK exposure and flag evaluation pipelines #12842
5. Add dd-openfeature SDK OpenFeature provider #12843 👈 this PR
6. Connect the dd-openfeature SDK to the agent #12844
7. Add agentless and standalone OpenFeature smoke tests #12846

Jira ticket: APMLP-1512

🤖 Generated with Claude Code

[datadog-prod-us1-4]

🎯 Code Coverage (details)
• Patch Coverage: 100.00%
• Overall Coverage: 59.83% (+0.01%)

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: 04c6c32 | Docs | Give us feedback!}

[dd-octo-sts]

🟢 Java Benchmark SLOs — All performance SLOs passed

Suite	Status
Startup	🟢 pass

SLO thresholds are defined here based on automatically generated metrics. A warning is raised when results are within 5% of the threshold.

PR vs. master results

Scenario	Candidate	master	Δ (95% CI of mean)
startup:insecure-bank:iast:Agent	14.02 s	13.90 s	[+0.0%; +1.7%] (maybe worse)
startup:insecure-bank:tracing:Agent	12.92 s	12.88 s	[-0.5%; +1.0%] (no difference)
startup:petclinic:appsec:Agent	17.08 s	16.98 s	[-0.3%; +1.5%] (no difference)
startup:petclinic:iast:Agent	16.74 s	16.98 s	[-2.0%; -0.8%] (maybe better)
startup:petclinic:profiling:Agent	16.89 s	16.74 s	[+0.1%; +1.8%] (maybe worse)
startup:petclinic:sca:Agent	17.05 s	16.85 s	[+0.3%; +2.0%] (maybe worse)
startup:petclinic:tracing:Agent	16.07 s	16.23 s	[-1.9%; -0.1%] (maybe better)

Commit: 04c6c32e · CI Pipeline · Benchmarking Platform UI

---

Load and DaCapo benchmarks can be triggered manually in the GitLab pipeline. Results will appear in the Benchmarking Platform UI after completion.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-16T19:39:58.722880Z	2d1dcc6	Draft marked ready
🔒 Security Review	✅ Completed 2026-10-16T19:35:43.618830Z	2d1dcc6	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2d1dcc6561

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[datadog-prod-us1-4]

Shutting down the last provider can discard buffered exposures from successful evaluations. Initialization can also exceed its configured timeout and block concurrent shutdown while waiting for configuration.

_{Open Bits AI session}

_{🤖 Bits Code Review · Commit 2d1dcc6}

[PerfectSlayer]

Pushed 2 other minor fixes and resolved thread where reports are not a regression from master.
Will proceed with merge. Thanks for the reviews!

[PerfectSlayer]

/merge

[gh-worker-devflow-routing-ef8351]

View all feedbacks in Devflow UI.

2026-10-20 08:40:47 UTC ℹ️ Start processing command /merge

---

2026-10-20 08:40:53 UTC ℹ️ MergeQueue: Pull request is not mergeable yet

It will be processed automatically as soon as GitHub reports it as mergeable. View in MergeQueue UI.

• Run /code blockers to see what is blocking it.
• Run /remove to cancel it.

---

2026-10-20 09:27:17 UTC ℹ️ MergeQueue: merge request added to the queue

The expected merge time in master is approximately 1h (p90).

---

2026-10-20 10:31:20 UTC ℹ️ MergeQueue: This merge request was merged