Skip to content

Fix native audit citations for separate supporting documents - #1918

Merged
Paul Lizer (paullizer) merged 1 commit into
microsoft:Developmentfrom
paullizer:paullizer-nasa-partial-results
Nov 11, 2026
Merged

Paul Lizer (paullizer) merged 1 commit into
microsoft:Developmentfrom
paullizer:paullizer-nasa-partial-results

Conversation

@paullizer

Copy link
Copy Markdown
Collaborator

Summary

  • Fix native document reviews that successfully read both a selected native file and a separate supporting document but return only a partial-result explanation because the supporting citation is excluded.
  • Supply an explicit retained-owner answer-citation registry while keeping native fact provenance, exact markers, version fences, and semantic checking strict. Advance the private checking contract to v4 so an explicit retry renews obsolete checking state without repeating completed reads.
  • Add real mixed-source HTTP/retention, correction, access-refusal, and recovery regressions, document the fix, and bump the application patch version to 0.261.428. No prompt heuristics, unchecked-answer fallback, or larger correction allowance.

Linked issue

N/A - no linked issue was supplied.

Release Notes & Latest Features

  • New Feature
  • Bug Fix
  • UI Enhancement
  • Breaking Change
  • Internal only

Is this visible to end users?

  • Yes
  • No

Is this admin-facing (Admin Settings, governance, deployment, config)?

  • Yes
  • No

Should this become a Latest Feature card?

  • Yes
  • No
  • Already added

Screenshot needed for the card?

  • Yes
  • No
  • Attached

Release notes are intentionally unchanged at the user's request. No new admin control or Latest Feature card is introduced.

Version bump

  • application/single_app/config.py VERSION third segment bumped, or not needed because this is docs-only
  • deployers/version.txt bumped, or not needed because deployers/ was not changed

Application version: 0.261.427 -> 0.261.428. No deployer files changed.

Testing / validation

Local validation used the existing network guard and isolated external I/O/provider replies. Counts below overlap and are not a combined unique-test total.

Passed: 218 tests covering native contracts/delivery, the real action HTTP path and lineage, completeness, and available-information controls:

$env:PYTHONDONTWRITEBYTECODE='1'
$env:PYTHONPATH="$PWD\application\single_app;$PWD\functional_tests"
python -B -m pytest .\functional_tests\test_native_audit_reliability.py .\functional_tests\test_native_audit_answer_delivery.py .\functional_tests\test_orchestration_document_action_end_to_end.py .\functional_tests\test_orchestration_document_action_lineage.py .\functional_tests\test_task_completeness_engine.py .\functional_tests\test_task_completeness_runtime.py .\functional_tests\test_orchestration_available_information.py -q --tb=short -p test_support.cosmos_network_guard -p no:cacheprovider --disable-warnings

Passed: 127 tests under optimized Python, including source-backed native pipeline and request-contract compatibility:

python -O -B -m pytest .\functional_tests\test_native_audit_reliability.py .\functional_tests\test_native_audit_answer_delivery.py .\functional_tests\test_native_document_fidelity_pipeline.py .\functional_tests\test_native_inspection_request_contract.py -q --tb=short -p test_support.cosmos_network_guard -p no:cacheprovider --disable-warnings

Passed: 38 tests for the final application version and focused citation contracts:

python -B -m pytest .\functional_tests\test_native_audit_reliability.py .\functional_tests\test_native_audit_answer_delivery.py::test_native_audit_delivery_version -q --tb=short -p test_support.cosmos_network_guard -p no:cacheprovider --disable-warnings
  • python -B .\scripts\check_cosmos_response_boundaries.py - passed, 1,094 owners, 425 doubles, zero review gaps.
  • python -B .\functional_tests\test_docs_app_surface_coverage.py - passed, 7/7 checks; inventory is current.
  • python -B .\functional_tests\test_docs_site_quality.py - passed, 6/6 checks.
  • git --no-pager diff --check - passed.
  • The previously failing separate-supporting-file HTTP cases now save checked answers; same-document controls still pass. Invented markers and revoked retained-action access still prevent publication. Explicit v3 recovery preserves native-download and supporting-search counts.

No live beta deployment or retry was performed. The saved trace establishes citation-validation failure, but the exact rejected markers in its private initial/final candidates were not recovered. Controlled regressions establish this contract repair, not guaranteed behavior from every live model. Retained workspace sources remain provenance; existing conversation/run and current external-action access semantics are unchanged.

Documentation

  • Release notes updated, or not needed
  • Feature documentation updated, or not needed
  • Fix documentation updated, or not needed

Added docs/explanation/fixes/NATIVE_AUDIT_MULTI_SOURCE_CITATION_FIX.md and linked it from the preceding reliability fix. No new feature surface requires a guide or inventory change. Release notes were explicitly left unchanged by the user.

Security checklist

  • New Flask routes include @swagger_route(security=get_auth_security())
  • Settings sent to non-admin frontends use sanitize_settings_for_user()
  • Browser JavaScript is served from local SimpleChat static assets only; no CDN-hosted JS
  • No secrets, keys, connection strings, or local-only artifacts are included

No routes, frontend settings payloads, or browser assets changed. Private investigation exports, console logs, and session artifacts are not included in this PR.

Admit supporting-document citations through the retained owner's projection while keeping native facts and strict marker validation separate. Version the checking contract for explicit renewal and cover mixed-source delivery, refusal, and recovery. Update fix documentation and application version to 0.261.428.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
@paullizer
Paul Lizer (paullizer) merged commit 2fafbd6 into microsoft:Development Nov 11, 2026
12 of 13 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant