Repository navigation
Validate final routing policies and harden artifact loading - #89
Merged
Merged
Conversation
adrida
force-pushed
the
fix/policy-validation
branch
4 times, most recently
from
September 29, 2026 17:54
31502d5 to
8de81bc
Compare
Reserve independent final-policy data, validate required OOD artifacts and stage artifact publication with rollback on exceptions. Update API documentation, examples and diagnostic reports to match serving behavior.
adrida
force-pushed
the
fix/policy-validation
branch
from
September 29, 2026 18:06
8de81bc to
2281840
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Check the selected routing policy once on untouched data, including its OOD filter, before publishing it. Reject missing or corrupt required OOD artifacts and preserve the previous generation if fitting or publication raises an exception.
A completed fit that does not pass certification publishes a null-policy manifest. The documentation explains that behavior, separate-directory promotion, teacher agreement, and the existing prediction API. Examples and scan-report wording are corrected accordingly.
The README introduces task-specific ML students, direct label prediction and teacher fallback before the API details. It distinguishes acceptance-policy deferral from OOD rejection, while preserving the two-value response schema, independent final-policy check and statistical limitations.
Validation: 71 tests passed with two optional skips; 12 final-policy safety checks and 11 scan/report checks passed again. Documentation examples, API arguments and local links were checked. The scan algorithm and prediction response schema are unchanged.