Skip to content

2.8.7.md

Latest commit

 

History

History
20 lines (15 loc) · 3.36 KB

2.8.7.md

File metadata and controls

20 lines (15 loc) · 3.36 KB
Raw
summary find_eligible age bounds and the enrollment sentinel filter no longer exclude studies with a missing field; sort rejections reclassified and normalized; query.* parameter descriptions completed; empty-list rejection extended to nctIds and get_field_values
breaking false
security false

2.8.7 — 2026-08-18

Fixed

  • clinicaltrials_find_eligible age bounds no longer drop studies that omit MinimumAge/MaximumAge — each bound is now ORed with its MISSING counterpart, since an AREA[Field]RANGE[…] predicate previously matched only studies publishing that field. Recruiting matches for a 58-year-old with Type 2 Diabetes go from 406 to 744 nationally. (#105)
  • The enrollment sentinel filter no longer excludes studies with no EnrollmentCount at all — restated as NOT AREA[EnrollmentCount]RANGE[99999999, MAX] instead of a bounded range, which had the same field-must-be-present gap as #105 and answered zero for every expanded-access record. clinicaltrials_search_studies also lifts the exclusion when nctIds is supplied, so an explicit ID filter can no longer omit a study it names; sentinelFilterActive now reports the exclusion that actually ran. (#106)
  • Three more upstream sort rejection shapes are now classified as sort_invalid instead of falling through to a generic 400 — the 2-item cap, an unsortable field type, and an unrecognized field name each get their own message, the last with did-you-mean suggestions. (#107)
  • Empty-list inputs ([]) are now rejected with blank_value instead of silently widening or falling back — .min(1) is removed from every list-typed parameter across search_studies, get_study_count, get_study_results, get_field_values, and find_eligible, each paired with a handler-level guard so the rejection carries a reason and recovery hint instead of a bare -32602. search_studies.nctIds: [] previously widened the query to the whole registry rather than being rejected. minItems no longer appears in any tool's advertised inputSchema. (#109, #110)

Changed

  • sort auto-corrects unambiguous casing and spacing mistakes before the request goes upstream — enrollmentCount:desc, EnrollmentCount:DESC, and a stray space after a comma are now normalized rather than rejected. (#107)
  • The seven query.*-backed parameter descriptions on search_studies and get_study_count now name the fields they actually match, sourced from a live fetch of /studies/search-areas; docs/api-reference.md corrects the query field count from "50+" to 57. (#108)
  • clinicaltrials_get_field_values rejects an empty fields list with reason blank_value instead of field_invalid — field_invalid's recovery hint pointed callers at browsing the field tree, which doesn't fit a caller who supplied []. (#109)