RELEASE_NOTES_5.1.0

Moira 5.1.0 — Source-Scoped Pancha Pakshi and Explicit Dignity Policy

Release date: 2026-07-21

Public upgrade path: 5.0.0 to 5.1.0

Moira 5.1.0 adds a first-class, source-scoped Pancha Pakshi subsystem and makes Halb and Oriental/Occidental dignity inclusion independently selectable. It also strengthens bulk eclipse-range execution, adds public Gauquelin validation evidence, and closes several release-validation gaps discovered while preparing the package.

The release remains governed by explicit policy and provenance. Pancha Pakshi does not select a universal canon, invent a numerical score, or turn research conflicts into prognostic truth. Dignity switches control the inclusion of already computed conditions; they do not alter the underlying astronomical or sect geometry.

Highlights

• Source-scoped Pancha Pakshi is available through immutable engine vessels, package-root exports, Moira facade methods, and typed /v1/pancha-pakshi REST routes.
• Fixed-clock and solar-proportional timing remain separate, explicitly named policies. No Pancha Pakshi profile is selected by default.
• Uromarisi constitutional evidence is inspectable across activity, classification, relation, condition, aggregate, network, and public-status layers without collapsing documented conflicts.
• Dignity callers can independently include or exclude Halb, Hayz, and Oriental/Occidental conditions through engine and REST policy vessels.
• Bulk solar and lunar eclipse ranges can use conservative native TT candidate discovery while Python retains time policy, physical refinement, classification, inclusive-range semantics, and result assembly.
• The g5 Gauquelin validation harness now carries explicit historical-dataset scope and validation evidence.

Pancha Pakshi public boundary

The admitted public profile is agastya_madras_1879_akshara_fixed_clock. Its authority is limited to the source product and locators recorded in its manifest. The API exposes profile discovery, aksara identity, schedules, source-scoped mappings, temporal selection, materialization, current-cell selection, relations, conditions, aggregates, networks, and constitution status.

The following boundaries are deliberate:

• profile selection is explicit and fails closed;
• fixed-clock and solar-proportional composition are different policies;
• exact rational nazhigai values remain fractions in Python and serialize as numerator/denominator pairs;
• research witnesses and conflicting cells retain their own admission status;
• no generic good/bad score, forecast, or empirical efficacy claim is added.

Dignity policy controls

DignityComputationPolicy now reaches independent accidental-condition controls:

• SectHayzPolicy.include_hayz;
• SectHayzPolicy.include_halb;
• AccidentalDignityPolicy.include_oriental_occidental.

All default to True, preserving 5.0.0 output. Setting one to False removes that condition's label, structured truth vessel, and additive contribution. It does not mutate planetary phase, sect, house placement, or longitude.

The same policy is accepted by all six dignity REST routes. The quarantined Valens distribution option remains unavailable rather than being exposed as an inert switch.

Astronomical and validation hardening

find_phenomena() now treats a missing meridian crossing inside its exact 24-hour search window as an absent event key. This matters for the Moon, whose successive meridian crossings can be separated by more than 24 hours. get_transit() remains fail-closed when its requested window contains no bracket, so direct callers retain explicit failure semantics.

Release preparation also aligned the physics-layer node reference with the reader-owned ephemeris clock, removed a false repository-wide prohibition on supported annotation syntax, made kernel-backed property tests independent of cold-cache timing, restored the session kernel after a state-mutating manifest test, and refreshed one deterministic DE441 regression fixture affected by the admitted clock correction.

Evidence boundaries

Pancha Pakshi source-table tests and constitutional invariants establish source fidelity and structural coherence, not empirical astrological efficacy. DE441-backed tests exercise astronomical routing and time conversion; they are not historical Pancha Pakshi oracles. Gauquelin historical-dataset tests are bounded to their named corpus. Snapshot and fixed-value tests remain regression evidence rather than external authority validation.

Publication

Pushing the v5.1.0 tag triggers the repository's GitHub Actions publication workflow. Its cibuildwheel matrix produces the supported platform and Python wheels and publishes them to PyPI. Local wheel production is not part of the release procedure.