Repository navigation
Align Codex target probe bytes and continuation cookie order - #275
Merged
Merged
Conversation
HTExplicit
enabled auto-merge (squash)
October 10, 2026 06:00
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Target STATE observations still use a generic map serializer and append the target's continuation cookie after the borrowed cookie. The pinned ranxi probe uses a fixed JSON byte template and retains the target cookie before appending the pinned borrowed cookie.
Use that fixed template and cookie order for the two-shot target protocol, while preserving the selected actual model, effective tier, per-shot legacy session IDs, plaintext encoding and strict STATE verdict. Business bodies, business cookie headers and variable Pelican generation payloads remain under their existing paths. Run the legacy sorted-body comparison first so later upstream failures are not always associated with that serializer.
Validation: exact probe body bytes and continuation cookie header, both actual models and native identity/tier preservation, fixed-candidate comparison isolation, encoding comparison, unchanged normal-generation compression and native HTTP tool continuation pass. Live tests have included a successful pinned-template pair followed by an inconclusive later pair; that is not sufficient to attribute failures to field order, and this alignment is not presented as a completed live repair.
Required CI exposed a pre-existing test cleanup race: Grok scheduler tests replaced a process-wide sync.Map while an asynchronous refresh from another test was storing into it, panicking inside sync.Map.Store. Those scheduler fixtures already own separate caches, so the unrelated global resets are removed. Production Grok behavior and existing assertions are unchanged; the directly affected scheduler tests are verified narrowly.
The required backend job also encountered ECR Public rate limiting before tests started. PostgreSQL and Redis preloading now uses the same fixed OCI digests through Google's Docker Hub cache, already used for builds. Both cache manifests returned the exact existing digests. Local image names, image contents and test commands stay unchanged; no credentials or client identities are rotated.