PM-5203 approval phase ai only

[vas3a]

This pull request introduces several important changes to improve the handling and reliability of AI Review and Approval phases in the challenge workflow. The main focus is to ensure that phases are only closed when all necessary AI decisions are finalized, and to correctly handle the Approval phase for AI-only challenges. Additional improvements include more robust passing score logic and enhanced test coverage.

AI Review and Approval Phase Handling:

• Added logic to block closing the AI Review phase if there are any pending AI decisions for active contest submissions, both in ChallengeApiService and SchedulerService, ensuring that all AI decisions are finalized before phase closure. [1] [2] [3] [4]
• For AI-only challenges, the Approval phase now skips human reviewer checks and closes on schedule, as there are no human reviewer scorecards required. [1] [2]

Review and Passing Score Logic:

• Improved the logic for determining passing scores: for submissions with a HUMAN_OVERRIDE status, the code now checks if the score meets the minimum passing threshold instead of always marking as passing. [1] [2] [3] [4]

Challenge Finalization Logic:

• In ChallengeCompletionService, finalization is now deferred if the Approval phase is still open or incomplete, ensuring winners are only determined after all manager overrides are applied. Also, finalizeSummations is now called after these checks. [1] [2] [3]

Testing Improvements:

• Added comprehensive tests to verify that the AI Review phase cannot be closed with pending AI decisions, and that non-AI phases and phase openings are unaffected.

Other:

• Added a new deploy:dev script to package.json for streamlined development deployments.

[Copilot]

Pull request overview

This PR tightens the AI Review/Approval phase lifecycle in the challenge workflow so phases don’t close (and AI-only challenges don’t finalize) until required AI decisions and approval conditions are satisfied.

Changes:

• Block closing the AI Review phase when there are pending AI decisions, in both the API-driven phase advance path and the scheduler-driven closure path.
• Adjust AI-only Approval phase closure behavior to skip human reviewer coverage/completion checks and close on schedule.
• Refine AI decision passing logic by applying minPassingThreshold checks for HUMAN_OVERRIDE, and add tests around the new AI Review closure blocking.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/review/review.service.ts	Adds pending-AI-decision counting and updates AI decision summary logic to incorporate minPassingThreshold.
src/challenge/challenge-api.service.ts	Blocks API phase closure for AI Review when pending AI decisions exist (via a new DB query helper).
src/challenge/challenge-api.service.spec.ts	Adds unit tests for the AI Review closure blocking behavior.
src/autopilot/services/scheduler.service.ts	Extends scheduler phase-closure gating for AI Review (pending decisions) and adjusts Approval closure rules for AI-only challenges.
src/autopilot/services/challenge-completion.service.ts	Defers AI-only finalization until Approval is closed and reorders summation finalization calls.
package.json	Adds a deploy:dev script for pushing a branch-based dev-* tag.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.