Issues

• Enforce the remote response ceiling structurally instead of by source-text match

  openmrs/openmrs-module-chartsearchai#627 · Filter by author pmanko opened on Oct 29, 2026
  1

• Grounding judge marks a verbatim restatement unsupported when the cited record is clinically implausible

  bug
  openmrs/openmrs-module-chartsearchai#612 · Filter by author dkayiwa opened on Oct 14, 2026
  2

• A question saying a drug was stopped gets no verdict on the drug it proposes: the stop is read as a current medication, and the ended-order template fires on the question's word

  openmrs/openmrs-module-chartsearchai#577 · Filter by author dkayiwa opened on Sep 30, 2026

• Deferred from #505: derived-tier coverage, ended-order referent gaps and the fold prompt experiment

  openmrs/openmrs-module-chartsearchai#521 · Filter by author dkayiwa opened on Sep 24, 2026
  1

• Every drug a question lists as current is screened as a proposal: her own orders' findings carry "a reason to withhold it", crowd the chip list ahead of the drug asked about, and on E2B decide its verdict

  openmrs/openmrs-module-chartsearchai#513 · Filter by author dkayiwa opened on Sep 24, 2026
  2

• The bundled KB's Sulfamethoxazole row carries sulfamethazine's RxCUI, RxNorm name and CIEL codes, so a cotrimoxazole order is shown to the clinician as "sulfamethazine" and the validity report does not count it

  openmrs/openmrs-module-chartsearchai#476 · Filter by author dkayiwa opened on Sep 23, 2026
  1

• Streaming AI query delivered to user but never audited when client closes connection early

  security
  openmrs/openmrs-module-chartsearchai#450 · Filter by author dkayiwa opened on Sep 16, 2026

• Clause-scoped citation split allocates O(markers x length) copies of an attacker-length answer

  security
  openmrs/openmrs-module-chartsearchai#448 · Filter by author dkayiwa opened on Sep 16, 2026

• Two WARNs quote the model's own off-schema citations field verbatim; Decision 102's sweep criterion reaches them

  openmrs/openmrs-module-chartsearchai#442 · Filter by author dkayiwa opened on Sep 16, 2026
  3

• Is an ATC code selected by the patient's own prescription a CWE-532 disclosure? Decision 102's residue, two shapes on two reasonings

  openmrs/openmrs-module-chartsearchai#441 · Filter by author dkayiwa opened on Sep 16, 2026

• Deploy reports success while the SPA serves the app a stale chartsearchai frontend bundle

  openmrs/openmrs-module-chartsearchai#434 · Filter by author dkayiwa opened on Sep 15, 2026
  2

• Two residues in the dosing-ceiling needle rule, from #424's review round

  openmrs/openmrs-module-chartsearchai#425 · Filter by author dkayiwa opened on Sep 14, 2026
  2

• DrugOrderRecords.numbersFor hands the uuid index's own List out of the class, where the name leg returns a fresh one

  openmrs/openmrs-module-chartsearchai#415 · Filter by author dkayiwa opened on Sep 13, 2026
  1

• An interaction screen across her own orders cites ten of the twenty findings it raised, and the completeness gate cannot tell that cell from one the module never asked to enumerate

  openmrs/openmrs-module-chartsearchai#399 · Filter by author dkayiwa opened on Sep 9, 2026
  1

• Proposal: consume the knowledge base's drug-disease rows and its derived tier (schema 1.2)

  enhancement
  openmrs/openmrs-module-chartsearchai#391 · Filter by author pbiondich opened on Sep 8, 2026
  2

• Nothing purges the CDN on deploy, and the bundle path never changes, so visitors were served a four-day-old frontend after a green deploy

  openmrs/openmrs-module-chartsearchai#390 · Filter by author dkayiwa opened on Sep 8, 2026
  1

• Dockerfile.frontend never invalidates its ESM clone, so rebuilding the frontend image ships the previous commit

  openmrs/openmrs-module-chartsearchai#389 · Filter by author dkayiwa opened on Sep 8, 2026
  1

• A finding names the order a substance came from by display string, so the model must join it to a record number itself — and mis-cites 3 of 5

  openmrs/openmrs-module-chartsearchai#379 · Filter by author dkayiwa opened on Sep 6, 2026
  3

• Tracking: the DDI report reproduction — coverage map, verification rig, and why #353-#358 will not make the report pass

  openmrs/openmrs-module-chartsearchai#359 · Filter by author dkayiwa opened on Sep 1, 2026

• No finding can state management: the KB has no management field, and the upstream tag survives only as the RECOMMENDED:/INTERVAL: residue #116 strips

  openmrs/openmrs-module-chartsearchai#358 · Filter by author dkayiwa opened on Sep 1, 2026
  1

• A question naming a drug class ("an oral contraceptive", "an NSAID") resolves to no substance, so no screen runs — while every member of the class interacts with a drug the patient is on

  ddi-automation-candidate
  openmrs/openmrs-module-chartsearchai#354 · Filter by author dkayiwa opened on Sep 1, 2026
  2

• Cotrimoxazole resolves to nothing: the KB's CIEL bridge keeps one name per code, so 15 of concept 105281's 16 names are dropped and a routine HIV co-prescription is never screened

  ddi-automation-candidate
  openmrs/openmrs-module-chartsearchai#353 · Filter by author dkayiwa opened on Sep 1, 2026
  1

• The answer paraphrases a deterministic safety finding and deletes the hazard: a Major methemoglobinemia warning reaches the clinician as "can also increase the risk"

  ddi-automation-candidate
  openmrs/openmrs-module-chartsearchai#337 · Filter by author dkayiwa opened on Aug 30, 2026
  5

• A stopped drug order can be reported without its stop date, and on one question shape the answer denies the status its own cited record states

  openmrs/openmrs-module-chartsearchai#315 · Filter by author dkayiwa opened on Aug 25, 2026
  4

• The record now carries the stricter presentation ceiling, but the answer still leads with the canonical row's

  enhancement
  openmrs/openmrs-module-chartsearchai#276 · Filter by author dkayiwa opened on Aug 18, 2026