Personalized Muhurta repair: validation and source-publication receipt

Date: 6 October 2026. Decision: VED-004 and VED-006 are locally complete for the bounded existing-profile, sampled/JD-weekday contract. The user subsequently authorized Git source publication of this package. Release, deployment, full classical authority validation, predictive validity, exact transitions and sunrise-owned search remain outside this receipt.

The owning standard defines the object, selected policy, evidence, compatibility changes and exclusions.

Runtime, scope and files

• Checkout C:\dev\moira, branch main, starting commit cbb4abcbd74e1ae5bf09c87ff3ee7006e620e471; initially clean.
• Project .venv\Scripts\python.exe: Python 3.14.3, Moira 6.9.9, moira/_moira_native.cp314-win_amd64.pyd loaded. No native numerical or build files were changed, so no rebuild was needed.
• tests/KNOWN_ISSUES.yml was empty. Selected tests used strict expiry, deny-by-default network policy and marked loopback HTTP test clients.
• Planetary resource was discovered with MOIRA_NO_DOWNLOAD=1 through find_planetary_kernel(): the installed user-cache de441.bsp. Harness content probes confirmed DE441. No kernel download or external network test was used.

File group	Change
moira/muhurta.py	Seven-weight/overflow validation, fixed reserved flag, named existing profile, strict natal inputs, repaired scalar and tuple adapters; generic rule tables and score coefficients retained
moira/muhurta_search.py	Frozen inspectable moment/window/search vessels, explicit reader clocks and true frames, typed resource/coverage failures, complete capped grid, threshold runs and stable ranking
moira/panchanga.py	Existing limb arithmetic extracted into one shared private reducer; public instant arithmetic retained
Root, facade, vedic, _facade_vedic	Eight shared-identity exports, two reader-bound methods, UTC/UT1 chart adaptation and sanitized reader failure
Muhurta/Panchanga server models and services, new search models/service	Strict numbers/timestamps, actual policy receipts, effective ayanamsa frame coherence, typed canonical search transport
Muhurta router, error mapping, OpenAPI/model exports	POST /v1/muhurta/search, specific 422 coverage/503 resource envelopes, discoverable classical-vedic metadata
Unit/integration/server tests	Hostile input, composition, reader, window, cap, export and HTTP contracts
Standards, initial transport design, API/REST reference, work register, Home, changelog	Current admission and remaining boundaries; June design marked historical; generated wiki synchronized through its owning tool

No other Vedic family, website, Urania, kernel, classical table, scientific constant, golden/snapshot artifact or numerical acceptance tolerance was modified. The full repository test suite and repository-wide lint were not run; the selections below own this receipt.

Verified evidence

Contract	Evidence and limit
Existing judgment	Existing generic/Tara/Chandra tests remain green, including nine-Tara, twelve-house Chandra, Chandrashtama and one-ulp Nakshatra cases. These are implementation/structural checks, not new edition collation.
Policy and numeric boundary	Seven weights reject bool/string/nonfinite/negative/overflow input; zero/custom weights preserve engine/HTTP results. Reserved flag changes reject instead of being ignored. Moment/chart numbers and numeric timestamps are strict.
Effective frame	Nested Panchanga policy owns the generic score and both personalized overlays. Generic/classification receipts disclose unapplied natal weights and the fixed rule profile.
Reader and clock	Explicit reader and context restoration checked on success/coverage failure; missing facade reader cannot borrow an ambient kernel. ChartContext UT1/TT and public Chart civil-UTC/Delta-T ownership are handled explicitly. Invalid dates and sample caps preflight before search resource access.
Installed ephemeris	All 12 named ayanamsas at two 6 October 2026 samples; 1600, 1900, 2000 and 2100 clock cases. Sun/Moon longitudes and ayanamsa compare with the existing owning engine using absolute 1e-10 degree tolerance; clock coordinates/receipts and personal scores agree exactly. This is shared-engine composition consistency, not independent JPL/IAU or external-engine validation.
Resource semantics	Actual DE441-uncovered epochs yield coverage errors; injected missing kernel/live-anchor cases yield resource failures. A missing live anchor does not use a polynomial fallback. Failures never yield a partial/empty success.
Sample selection	Both endpoints exactly once, clipped final step, exact 4096 budget including endpoint, rejected-point splits, singleton-capable spans, earliest peak ties, later stronger run retained before top cap, adjacent brackets, successful empty selection and explicit truncation. Synthetic scan fixtures are not astronomical source fixtures.
Legacy compatibility	Named natal Nakshatra without Moon rejects, contradictory natal inputs reject, supplied ElectionalEvaluation unwraps its tropical chart, incompatible legacy body/frame/gap/refinement settings reject, and the tuple view matches the canonical personal search.
HTTP	Installed-engine result parity for fields/order/peak/score/clocks, hostile scope rejected before compute, structured 422/503/request-ID behavior, GET rejection, 12-ayanamsa and cap OpenAPI schemas, and discovery family.
Exports/discovery	Root/facade/Vedic share identity; static public/method inventory passes. The earlier published dated-Gochar eleven names/two methods were explicitly added to the stale static guard. Gochar tag ordering was reconciled to the existing metadata/core-first test; no Gochar runtime behavior changed.

Commands and receipts

All test commands were prefixed with:

$env:MOIRA_TEST_MODE = "1"
$env:MOIRA_STRICT_KNOWN_ISSUES = "1"

The broad final selection:

.venv\Scripts\python.exe -m pytest tests\unit\test_muhurta.py tests\unit\test_muhurta_search.py tests\unit\test_panchanga.py tests\unit\test_daily_panchanga.py tests\unit\test_electional.py tests\unit\test_vedic_surface_completeness.py tests\unit\test_api_surface_adversarial_audit.py tests\integration\test_muhurta_search.py tests\server\test_server_muhurta_search.py tests\server\test_server_muhurta_routes.py tests\server\test_server_panchanga_routes.py tests\server\test_server_panchanga_service.py tests\server\test_server_vedic_phase2_routes.py tests\server\test_server_route_discoverability.py tests\server\test_server_error_mapping.py -m "not external_network" -q --junitxml=C:\dev\outputs\muhurta-personal-search-2026-10-06\targeted.xml

Result: 630 passed, zero failed/errors/skips, 91.107 seconds. Harness: 111 successful planetary-resource receipts, one DE441 content probe.

The final chart/cap/HTTP slice after explicit civil-chart adaptation and date bounds:

.venv\Scripts\python.exe -m pytest tests\unit\test_muhurta_search.py tests\integration\test_muhurta_search.py tests\server\test_server_muhurta_search.py -q --junitxml=C:\dev\outputs\muhurta-personal-search-2026-10-06\final-search.xml

Result: 143 passed, zero failed/errors/skips, 30.991 seconds; 81 successful resource receipts and one DE441 probe.

Final reader/preflight slice after sanitized facade resource handling:

.venv\Scripts\python.exe -m pytest tests\unit\test_muhurta_search.py tests\integration\test_muhurta_search.py -q --junitxml=C:\dev\outputs\muhurta-personal-search-2026-10-06\final-reader.xml

Result: 87 passed, zero failed/errors/skips, 6.392 seconds; 23 successful resource receipts and one DE441 probe. The union of JUnit (classname, name) keys is 635 distinct passing cases, not the sum of repeated runs. XML attributes and the union were checked with the project runtime and saved to C:\dev\outputs\muhurta-personal-search-2026-10-06\summary.json.

The saved standard's executable example also ran against the installed engine: 25 reader-bound samples, one returned run, identity DE-0441LE-0441. Assertions verified exact start/end, TDB presence, personal mode, threshold qualification and peak membership. This was a real-kernel contract smoke.

Scoped lint and documentation checks:

.venv\Scripts\ruff.exe check moira\muhurta.py moira\muhurta_search.py moira\panchanga.py moira\_facade_vedic.py moira_server\models\muhurta.py moira_server\models\muhurta_search.py moira_server\models\panchanga.py moira_server\services\muhurta.py moira_server\services\muhurta_search.py moira_server\routers\muhurta.py moira_server\errors.py moira_server\openapi.py tests\unit\test_muhurta_search.py tests\integration\test_muhurta_search.py tests\server\test_server_muhurta_search.py --no-fix
.venv\Scripts\python.exe scripts\sync_rest_api_reference.py --check
.venv\Scripts\python.exe scripts\check_doc_consistency.py
.venv\Scripts\python.exe scripts\sync_git_wiki.py
.venv\Scripts\python.exe scripts\sync_git_wiki.py --check
git diff --check

All final checks passed. Python 3.10 grammar was also checked for the new module, models, service and affected facade/scoring/reducer files; other Python runtime versions were not executed. Local wiki synchronization uses intent-to-add entries for the two new canonical Markdown pages because the generator discovers tracked files. At that validation checkpoint no file content was staged, and no commit/push had been performed.

The initial strict-number assertion mismatch, historical tag ordering and stale export guard were resolved explicitly; no golden data or numerical tolerance was regenerated to make a test pass. There are no unresolved failures in the selected final receipts.

Source-publication scope, 6 October 2026

The subsequent instruction was "please commit and push these changes". The publication package contains the existing Muhurta engine/public/REST changes, their tests and canonical/generated documentation, plus the parent wiki gitlink. The nested wiki is committed and pushed before the parent engine. The engine remains version 6.9.9; no release tag, package publication or deployment is part of this operation.

The saved test receipts above remain the numerical and behavioral evidence. Only this receipt and the work-register publication wording were edited after those checks. Publication checks use the project runtime for REST inventory, documentation consistency and generated-wiki consistency, alongside scoped lint, whitespace checks, staged manifests and exact local/remote branch-head comparison. The containing source commits carry this publication package; their identities are reported in the publication completion message.