VED-012 execution and validation receipt

Date: 12 October 2026. Scope: all seven stages of the authorized implementation plan. Selected package is locally complete; no commit, push, tag, release or deployment was performed by this implementation run. Version remains 6.9.9. Starting engine revision: 5049a525f482dbfc60b5fde5001a39c034627f58.

Subsequent publication authorization and refreshed gates are recorded in the publication checkpoint below; the earlier implementation record is retained as historical evidence.

The delivered standard governs the 21 source rule entries, four explicit profiles, Python/facade/REST contracts and legacy migration. BPHS 39.35 is discoverable but non-executable.

Stage closure

Stage	Delivered evidence
1 — source admission	Immutable source/profile/rule catalogue; scan hashes and page locators; 44 independent cases and separately transcribed Raman Chart 122 signs
2 — legacy repairs	Shared admission for all public families/classifier; strict doctrine/transport booleans; missing-motion state, indeterminate formation and applied-policy transport
3 — Navamsa	Two distinct profiles, immutable call-local context, canonical D9 owner, frame and self-dispositor cases, exact binary boundary checks
4 — wealth/Raja	All fourteen wealth rows and five selected royal entries; all required supporters, roles and branch directions retained
5 — public/REST	Identical root/facade/Vedic exports, thin facade method, typed discovery/evaluation, schema example, source-corpus parity through actual ASGI HTTP
6 — validation/cost	573 passing selected tests; 144-chart legacy comparison; reproducible 132-request warm engine/HTTP measurements; normal production engine factory startup
7 — documentation	Standard, Python/REST references, fixture provenance, this receipt, executed plan and register; release-facing consistency and generated-inventory guards

Two implementation refinements avoid unnecessary work without changing semantics: D9 Lagna is not calculated because no selected rule uses it, and the seven-body context uses bounded direct sign-aspect predicates rather than precomputing an unused full matrix. Evaluated clauses have two truth states; no-candidate applicability belongs to the rule, and unevaluated interpretive context belongs to a typed qualification. This makes those distinct concepts explicit without introducing unsupported formation states.

Evidence classes and acceptance

Primary source / selected interpretation: the research packet and tests/fixtures/yoga_extensions/README.md identify JP/Sastri 7.14, Raman no. 252/Chart 122 and BPHS/Santhanam 41.2–15 and 39.33–39. The JSON fixture contains 44 manually derived positive/counterexample cases, with every executable ID represented. Expected states were not generated from the evaluator. Published Chart 122 D1/D9 signs are distinct from the synthetic degrees used to exercise the API. No historical birth ephemeris or predictive accuracy is claimed.

Independent geometry: all twelve ascendants of the H5 wealth tables; every named supporter separately; all twelve source signs for every supporter in both own-sign H1 variants; 2,688 royal-rule charts checked branch-by-branch against a separate sign/lord/aspect oracle. JP tests separate natal/D9 frames, D1 Lagna-lord identity, both movable gates and valid self-dispositor placement. Identity aliases do not create self-conjunctions.

D9 tests use rational arithmetic for all 108 partition starts and their adjacent representable values, 108 modality-defined interior positions, and the representable neighbor below 360 degrees. Results and discrete doctrine states require exact equality: no epsilon snapping or relaxed tolerances. BPHS 39.35 has the separate 84-case demonstration that no one full-graha source can reach both opposite H4/H10 targets.

Regression: 144 deterministic charts captured before legacy edits were compared field-by-field afterward, across twelve Lagna signs and doctrine variants. Zero differences in existing computed fields. Exactly 75 source strings received the intended JP “separately available” wording update. New evidence-state fields were additive and excluded only because the old schema did not contain them; booleans, ordering and suppression were compared. This is regression evidence, not a primary authority oracle.

Private capture: C:/dev/outputs/ved012-research-2026-10-12/legacy_before.json, SHA-256 fc69926a20938ecc95116a53d0730a3d8ea38d43331b71d63945d15a9facb5bd. Comparison receipt: sibling legacy_comparison.json.

Transport: all 44 cases run through defining module, root, facade, Vedic namespace, method, service and HTTP. Full source/role/frame/branch, state and policy payloads compare exactly. Tests cover filtering (including empty filtered results), complete counts, canonical profile ordering, concurrent request independence, model curation, OpenAPI, extra/missing inputs, invalid selections, wrong scalar types and non-finite values. The service revalidates deliberately forged request models; contradictory state/observation vessels fail at construction or serialization.

Legacy HTTP tests verify strict booleans, applied policy, retained indeterminate results and the distinction between absent speed, observed direct speed and retrograde speed. Existing yoga/varga/Vedic hardening and public-surface audit tests pass in the same run.

Executed commands and runtime

Project .venv, Python 3.14.3 on Windows 11. tests/KNOWN_ISSUES.yml contains no exemptions. Downloads disabled; external network denied by the test harness. HTTP tests use its loopback capability. No native/astronomical implementation or package dependency changed.

$env:MOIRA_NO_DOWNLOAD = '1'
$env:MOIRA_TEST_MODE = '1'
$env:MOIRA_STRICT_KNOWN_ISSUES = '1'
.\.venv\Scripts\python.exe -m pytest tests/unit/test_yoga_extensions.py tests/unit/test_yoga_extension_admission.py tests/unit/test_yogas.py tests/unit/test_varga_exact_partition.py tests/unit/test_api_surface_adversarial_audit.py tests/server/test_server_yoga_extensions.py tests/server/test_vedic_direct_input_hardening.py -q -rA --junitxml=C:/dev/outputs/ved012-research-2026-10-12/acceptance.xml

Result: 573 tests, zero failures, errors or skips, 50.209 seconds in the JUnit receipt. Normal unmocked engine creation ran through the FastAPI lifespan. Most contract cases use a deliberately minimal engine fixture; the separate lifecycle test establishes actual startup. The evaluator is resource-free; the application's ordinary startup still discovers and opens configured resources. The local performance run reported 471 ready readers. No-download discovery resolved the existing planetary kernel to C:/Users/nilad/.moira/kernels/de441.bsp.

After the final catalogue metadata and export edits, a second run of test_server_yoga_extensions.py, test_yoga_extensions.py and test_api_surface_adversarial_audit.py passed 307 tests, zero failures/errors/skips, in 41.560 seconds (final_transport.xml alongside the first receipt). These are repeated members of the 573-test selection, not 307 additional distinct tests.

Additional executed gates:

• Full configured Ruff rules pass for the 14 changed owner/service/test/tool files outside aggregate export modules. Aggregate modules pass the release gate's F401/F821/F822/F823 selection; their existing E402/F811 debt remains separate from this package.
• Docstring preflight passes for all eight changed engine Python files.
• Documentation consistency, release identity 6.9.9, REST generated route inventory and Hellenistic generated inventories pass. Generators were run; the Hellenistic content remained unchanged.
• git diff --check passes. The YAML adds this yoga slice as a release gate; remote CI was not run or claimed.

The repository-wide docstring preflight initially reported 23 MC-001 issues in untouched moira/houses.py under a concurrently edited governance checker. The eight-file VED-012 preflight passed. Other concurrent governance, event-search/Gochar and light-box work was preserved; no unrelated repair or broad green-CI claim is part of this receipt. Generated moira.wiki synchronization is reserved for the separately authorized publication step, as the plan specifies.

Cost and resource receipt

Reproduce with:

.\.venv\Scripts\python.exe tests/tools/benchmark_yoga_extensions.py --output C:/dev/outputs/ved012-research-2026-10-12/performance.json

Corpus SHA-256: 8f11e0b6485c1ee796ff3b7d545d793083a3eaf69eef34f2e4f0a91dce7b8be6. 44 source cases, all four profiles / 21 rules per request, three repetitions (132 warm calls for each path). Every measured HTTP response compares to the full canonical engine evidence. No clauses or validation cases are skipped to obtain the timings.

Measurement	Observed milliseconds
First engine call after imports	1.257
Warm engine median / p95 / max	1.103 / 1.646 / 1.888
Normal app creation and lifespan startup, prewarm off	25,627.117
First HTTP call after startup	12.283
Warm in-process HTTP median / p95 / max	6.100 / 8.154 / 72.427

JSON payloads: 86,642–93,599 bytes; median 86,739.5 bytes. Full branch and source evidence is deliberately retained. These are local observations, not latency guarantees or TCP/production benchmarks. Application resource startup is measured separately from the supplied-longitude calculation.

Instrumented tests establish seven D9 calls for all-profile evaluation, zero for wealth/Raja-only requests, and no calls to planetary evaluation, kernel discovery, reader acquisition or kernel opening during evaluation. The test harness denies external network. There is no global chart cache.

Remaining boundaries

BPHS 39.35 needs a cited resolution of aspect targets/type and a positive satisfiable source witness before a future admission. Other chapter-specific extensions and additional source traditions retain their separate scope. The delivered package evaluates selected formation, not strength/dasha, historical outcomes, natal interpretation, or date-derived yoga searches.

Authorized source-publication checkpoint — 12 October 2026

The user authorized committing and pushing VED-012. The refreshed checkout includes the separately published docstring-governance/house repairs through a399faa; the full docstring preflight now passes 366 engine Python files. The earlier 23-house-class report is resolved by that separate work.

The same seven-file acceptance command was rerun with its JUnit output at C:/dev/outputs/ved012-research-2026-10-12/publication.xml: 573 tests pass, zero failures, errors or skips. The 14-file scoped Ruff check and aggregate F401/F821/F822/F823 gate pass, as do documentation consistency, release identity 6.9.9, Hellenistic inventory and REST reference checks.

Publication support also passes 40 tests via python -m pytest tests/unit/test_git_wiki_sync.py tests/unit/test_public_api_drift.py tests/unit/test_public_api_contract.py -q using the same project interpreter and strict no-download test environment. Its JUnit receipt is publication-support.xml beside publication.xml. The generator synchronized seven intended wiki pages with no deletions; scripts/sync_git_wiki.py --check and both repositories' whitespace checks pass.

The source package comprises the VED-012 engine/public/REST changes, source fixtures and tests, benchmark tool, its release-hardening step, canonical documentation and generated Git-wiki mirror. The wiki is published before the parent engine commit records its gitlink. The engine commit containing this checkpoint and its moira.wiki gitlink identify the pair without a self-referential commit hash in either source document. Concurrent light-box scripts, tests and artifacts are excluded. No tag, package release, deployment or new VED-013 implementation is included.