Repository navigation
Fix sequential RSI monitoring and require cloud evaluation capacity - #121
Merged
w4ffl35 merged 4 commits intoOct 12, 2026
Merged
Conversation
Require fresh workers for mutation and CPU evaluation classes before baseline, campaign, or inference admission. Document separate and combined cloud worker configuration and cover stale registrations and missing evaluator capacity. Refs #120
w4ffl35
marked this pull request as ready for review
October 12, 2026 17:32
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
A paired RSI proof registers its fixed-control arm before its self-hosted arm. The operations monitor now reports absent arms explicitly instead of rejecting this valid sequential state. Unknown or duplicate arms still fail closed, and both arms must be finished before terminal accounting can be reported as settled.
Cloud RSI admission now requires fresh worker capacity for both REMOTE_API inference jobs and CPU evaluation jobs before baseline work or campaign admission. Previously a cloud-only worker could leave the first CPU evaluation queued indefinitely. Documentation covers separate or combined workers. CPU evaluation runs code and tests; it does not run a local language model. The version is prepared for v1.9.9.
Closes #119. Closes #120.
Verification
The controller integration fixture was supplied its 59 pinned historical benchmark objects. Earlier qualification failures from the missing objects and insufficient scratch-directory permissions are retained privately; the corrected runs changed the environment without changing source. No local inference or GPU testing was performed.
Security impact
The monitor remains read-only and retains all lease and accounting failure checks. Admission checks add an early refusal for missing worker classes without changing inference budgets, provider restrictions, or product GPU capability.
Test plan
Run npm run typecheck and npm test. Enable the optional PostgreSQL/S3 and isolated monitor test database variables to exercise the database checks; hosted CI runs these optional tests only when configured. Run bash scripts/e2e/run.sh for the CLI/HTTP mock checks.