Closed
[WIP] Harden authentication and protect sensitive tokens#60
Conversation
BigSimmo
marked this pull request as ready for review
June 23, 2026 05:55
BigSimmo
pushed a commit
that referenced
this pull request
Jul 21, 2026
…59) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EXsJcLrbZUXwnBeG91cVo9
BigSimmo
pushed a commit
that referenced
this pull request
Jul 21, 2026
…iss class) The monitoring-schedule answer gate was narrower than the fact filter and the eval judge: inflected schedules (reviewed annually, monitored for 3 hours), bare digit durations, and the level-range/metabolic-panel vocabulary passed fact filtering but died at the gate or at kind classification, so answers for those cases shipped without the asked-for schedule (run-#60 targeting misses). One shared monitoringScheduleEvidencePattern now feeds both the gate and the filter (lockstep by construction); the kind arm gains inflection tolerance only (panel/range vocab would steal sentences from the dose arm — pinned by the restored adjacent-context refusal test); and both the result-level and sentence-level coverage checks gain a figure escape mirroring the existing concrete-dose escape, tightly scoped to sentences/chunks that carry the asked-for interval or unit range themselves. Four red-proof tests (each verified failing pre-change) plus a negative guard pin the widen. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EXsJcLrbZUXwnBeG91cVo9
BigSimmo
pushed a commit
that referenced
this pull request
Jul 21, 2026
…up layer The neuroleptic-side-effect-escalation eval case failed the citation gate on every canary (#57-#60): the query misclassifies as medication_dose_risk, its honest lexical pool fails the dose fast-path floor (measured 0.184/0.0125 vs 0.66/0.055), the S3 document-lookup title layer that would surface the title-named SOP was excluded by the class allowlist, and the vector leg then injects the wrong sibling doc. shouldAttemptDocumentLookupFastPath now also engages the layer for medication_dose_risk when the classifier's existing deterministic signals say the query is escalation-shaped (intent escalation_risk — only assigned when drug_dosing wording did NOT match, so pure dose/route/frequency questions structurally cannot fire) AND a curated title alias phrase is present for the alias tier to rescue with (documentTitleTerms > 0). Rescue semantics by construction: the S3 block sits after both fast-path returns, so a medication_dose_risk query only reaches it once the floor already rejected the pool. Executed blast-radius sweep: exactly one eval case fires across the 44-case answer canary, the 36-case golden retrieval fixture, and the 30-case answer-quality harness — every non-firing case executes byte-identical code. No imputation formula, released-comparator key, selection clamp, or threshold is touched. Red-proof: the predicate test and the end-to-end rescue ordering test both fail with the gate reverted (verified); negative guards pin pure-dose, no-title, and non-escalation shapes plus the four allowlisted classes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EXsJcLrbZUXwnBeG91cVo9
13 tasks
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.
Thanks for asking me to work on this. I will get started on it and keep this PR's description up to date as I form a plan and make progress.