feat(v21): Learned Policy decision outcome closed loop

[laiqian0239-glitch]

Work package

V21-LEARNING-POLICY-P1-DECISION-OUTCOME-CLOSED-LOOP-V1

Authorized by ordinary two-parent merge 915c487784d991e5c683daa6e167950f310b9a47 from #371.

Failure-first topology

Exact first implementation commit: fb1beeb0e39ad4d6263f71b4c995eb5ec663e6e2

It is the direct child of the authorization merge and adds exactly the six authorized WP0 tests, with zero product/runtime/dependency implementation.

Causal RED evidence

Stage run 31796413179, job 94754440534 at exact Head fb1beeb0e39ad4d6263f71b4c995eb5ec663e6e2 reached the product route after the base-owned failure-first/implementation-branch contracts passed.

npm run test:wp0 result: 402 pass / 7 fail. The seven failures are exclusively the new Learned Policy contracts:

• missing backend/services/learningPolicyDecisionContract.js;
• missing backend/services/learningOutcomeAttributionService.js;
• production contextAwareReplyBrain.js does not yet consume learningPolicyRuntimeAdapter before existing frontier generation;
• existing learningDeepTrainingContract has no projectPolicy method;
• VW supply-chain closure is absent, including third_party/licenses/vowpal-wabbit-BSD-3-Clause.txt;
• missing backend/services/learningPolicyRuntimeAdapter.js.

All pre-existing WP0 files in the isolated runner passed; ACV2 run 31796413446 and WP-A post-merge run 31796413171 also succeeded on the test-only Head. This is the required causal product RED, not a runner or authorization failure.

Authorized GREEN boundary

Implementation remains restricted to the 22 exact paths frozen by #371. Mature Vowpal Wabbit 9.11.2 at exact upstream commit 122bae254a5b8bc2b774d13b33d53e6dbc2cfba7 owns the contextual-bandit action head. Existing Person/Persona/canonical serialization/immutable learning ledger/Langfuse reward/Model Brain provider-routing and frontier-generation authorities remain authoritative. P1 is deterministic (actionProbability=1.0, exploration=false) and may change only candidateStrategyBranch.

No workflow/root npm mutation, new general-purpose Yance infrastructure, live randomized exploration, local learned-policy reply generation, release/publish/promotion, or automatic scope expansion is authorized.

[coderabbitai]

Warning

Review limit reached

@laiqian0239-glitch, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 28 minutes

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: e880d88b-780e-43e4-9bfd-01b8335fca42

📥 Commits

Reviewing files that changed from the base of the PR and between 915c487 and fb1beeb.

📒 Files selected for processing (6)

• tests/wp0/v21-learning-policy-p1-decision-record.test.js
• tests/wp0/v21-learning-policy-p1-outcome-binding.test.js
• tests/wp0/v21-learning-policy-p1-production-consumption.test.js
• tests/wp0/v21-learning-policy-p1-projection.test.js
• tests/wp0/v21-learning-policy-p1-supply-chain.test.js
• tests/wp0/v21-learning-policy-p1-vw-runtime.test.js

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[laiqian0239-glitch]

Superseded by the ordinary-merged V3 production-closure authority and implementation. The merged V3 authorization explicitly records PR #372 / head df6143b4958401c8b1a4a6b24b29fd09e3909fbf as immutable audit evidence that is not mergeable, and forbids reuse/cherry-pick/history rewrite of superseded V1/V2 as executable authority. V3 implementation PR #384 was ordinary-merged to trusted main as 9252ebba53d0e6d4bd0388a88ede2d0e74c7164c. Closing this stale V1 PR without rebasing, rewriting, cherry-picking, or touching its audit history.

[laiqian0239-glitch]

Audit correction to the preceding closure note: the direct supersession authority for PR #372 is the ordinary-merged V2 successor authorization governance/layered-ci/v21-learning-policy-p1-decision-outcome-v2-successor-authorization.json, which records historicalImplementationPullRequest: 372, frozen head fb1beeb0e39ad4d6263f71b4c995eb5ec663e6e2, historicalImplementationMayMerge: false, and requires the historical implementation to remain unmodified. V3 directly supersedes V2/#379, not #372. The closure decision is therefore unchanged, but this comment corrects the authority chain precisely.