moira

Pancha Pakshi Research And Admission Standard

Subsystem: moira/pancha_pakshi.py with private ingestion in moira/_pancha_pakshi.py

Computational domain: Source-scoped five-bird timing doctrine

Status: Four named source-scoped public profiles; no default or universal canon

1. Current Boundary

Pancha Pakshi is within Moira’s astrological domain, but the name does not identify one interchangeable computational canon. Inspected Tamil witnesses disagree about identity, durations, schedule cells, relationships, and the relation between verse and commentary. Moira therefore admits independently named products and never selects an ambient Pancha Pakshi default.

The first public profile is agastya_madras_1879_akshara_fixed_clock. Its public claim is limited to the Agastya-attributed Madras 1879 aksara/query-or-name-initial fixed-clock operating schedule and directed relationship matrix. It admits exactly these capabilities:

• aksara_identity;
• nominal_schedule;
• first_eat_bird_mapping;
• directed_relationships;
• astronomical_context;
• astronomical_paksha_inference;
• fixed_clock_materialization;
• fixed_clock_current_cell_selection;
• solar_proportional_materialization; and
• solar_proportional_current_cell_selection.

It is not a natal-Moon or birth-nakshatra profile. Nominal-schedule callers supply purva or amara, day or night, and weekday explicitly. The separate local-solar context operation accepts an instant and location, derives only the governing sunrise window, day/night half, and local-mean-solar weekday, and still requires the caller to supply purva or amara. The separate fixed-clock operation materializes the selected nominal offsets from the local-solar half start under an explicitly modern policy. A third modern operation selects one current fixed-clock cell only after resolving the governing solar half, and reports an explicit unmaterialized tail when a long solar half outlasts the fixed span. Two further modern operations proportionally map the unchanged nominal fractions across the complete governing solar half and select one current proportional cell. Stage 2F is a separate location-free astronomical inference: it maps apparent geocentric Moon-Sun phase halves to the profile’s directly attested Purva or Amara label. It does not select a schedule, supply paksha to another operation, or infer natal identity. Every schedule, materialization, and current-cell caller still supplies purva or amara explicitly.

Stage 2I exposes one already normalized source component of that same 1879 profile as a pure lookup: the named schedule generator’s first-samam EAT seed. It accepts explicit profile Paksha, day/night half, and weekday, and performs no schedule materialization, temporal routing, or interpretation as a whole-day eating bird, Padu, Adhikara/Bharana, authority bird, condition, or score.

Stage 2J recovered a separately witnessed vinadi research object without admitting a runtime capability. The bounded object is a five-position ordinal axis beneath each named activity. Uromarisi-attributed witnesses require accurate vinadi division without supplying arithmetic, while a Bogamuni-attributed editorial witness separately supplies two incompatible selectors: weighted Sūkṣma intervals and equal-fifths Eka Sūkṣma intervals. Neither selector is bound to the Uromarisi ordinal outcomes, and neither is a default. Stage 2K admits only the two Bogamuni editorial selectors through a separate profile and mandatory caller policy selection. No admitted profile, manifest capability, public vessel, facade method, or REST route may infer the cross-witness Uromarisi composition.

The second public profile is bogamuni_chennai_2024_nakshatra_natal_identity. It admits exactly two capabilities:

• nakshatra_bird_mapping; and
• natal_identity.

Its source product is a complete 54-cell Purva/Amara-by-nakshatra bird table. Its separate natal product is a fixed modern Moira composition over that table, not a claim that the source explicitly specifies birth-Moon or ayanamsa calculation. It supplies no aksara identity, schedule, relationship, materialization, current cell, condition, score, or forecast, and it does not alter the 1879 profile.

The third public profile is bogamuni_chennai_2024_padu_bird_mapping. It admits exactly one capability:

• padu_bird_mapping.

Its product is the source’s separate fourteen-cell Purva/Amara-by-weekday Padu bird table. Callers supply both source Paksha and weekday explicitly. The table has no day/night axis, performs no astronomical routing, and is not a schedule or identity product. Padu remains the source-labelled death-or-inoperative bird; it is not converted into RULE, first_eat_bird, an Adhikara bird, or a generic authority bird. The profile supplies no natal identity, schedule, materialization, current cell, condition, score, or forecast, and it alters neither prior profile.

The fourth public profile is bogamuni_chennai_2024_sookshma_temporal_selector. It admits exactly one capability:

• sookshma_temporal_selection.

Each call supplies an explicit selector policy, parent activity, and exact elapsed Fraction within one six-nazhigai samam. The weighted policy rotates the source-attested activity-duration vector from the parent activity. The equal-fifths policy yields five ordinal-only cells and assigns no subactivity. Both use exact half-open intervals. The profile supplies no datetime, astronomical context, schedule composition, Uromarisi outcome, condition, score, electional window, or forecast. It has no default, and no human-language reviewer is required.

Stages 2N and 2O add no fifth profile and change none of these capabilities. They are separate modern composition operations across the named 1879 schedule profile and the named 2024 selector profile. Every policy-driving profile, schedule or civil-time axis, subject bird, timing policy, selector policy, and explicit or derived exact elapsed offset remains visible.

2. Admission Tiers

Admission belongs to a named profile and product, not to Pancha Pakshi as a whole.

Status	Meaning	Public computation	Default selection
research_only	Incomplete, unresolved, or non-admitted witness work.	No	No
source_scoped_public	One named witness and declared text-layer policy, with limitations carried in every result.	Yes	No
corroborated_public	A public named profile with documented independent-witness corroboration; disagreements remain distinct.	Yes	No

There is no universal_default status. The schema rejects default_selection_allowed=true. Adding a default would require a new doctrinal and schema decision rather than an unnoticed manifest edit.

Machine-assisted source reading is admissible when the exact witness, rendered page locator, text-layer uncertainty, conflicts, and nonclaims remain visible. There is no human-language-review dependency. Independent-witness collation is required only for corroborated_public; it is not required for a narrowly and honestly labelled source_scoped_public product.

3. Governing Objects

The implementation keeps these objects distinct:

• witness — one identified edition or manuscript record;
• text layer — verse, commentary, printed table, or an explicit rule derived from those objects;
• profile — one named, internally coherent selection from one witness;
• product kind — for example, an aksara/prasna operating schedule rather than natal-bird timing;
• capability — an independently admitted computation within a profile;
• regime — Purva or Amara, separately for day and night;
• schedule cell — one bird, activity, samam, sequence position, exact duration, and source locator;
• identity policy — query/name initial, birth nakshatra, or another explicitly sourced or explicitly modern-composed identity product; and
• timing policy — source-attested fixed nazhigai timing or a separately named modern seasonal transformation; and
• local-solar context policy — a separately admitted modern composition that derives astronomical context without altering the source schedule or pretending that the composition is source-attested; and
• fixed-clock materialization policy — a separately admitted modern composition that binds a sourced fixed nazhigai unit to reader-bound TT, publishes UT1 intervals, and preserves the fixed schedule’s unclipped relation to its astronomical half; and
• fixed-clock current-cell selection policy — a separately admitted modern half-open interval-membership policy that resolves the governing solar half before selection and preserves an explicit unmaterialized long-half tail;
• solar-proportional materialization policy — a separately admitted modern composition that maps exact nominal offset fractions over the complete reader-bound TT solar half without using the fixed nazhigai-second rule; and
• solar-proportional current-cell selection policy — a separately admitted modern half-open interval-membership policy over that complete proportional partition, with exactly one non-null selected member and no tail or fixed-clock fallback; and
• astronomical paksha inference policy — a separately admitted modern geocentric lunar-phase classifier whose exact half-open numeric boundary is Moira-owned and whose waxing/Purva and waning/Amara translation is owned by the named source profile; and
• Padu-bird mapping - one source-owned (Purva|Amara, weekday) -> bird table with no day/night or schedule dimension and no implicit synonymy with authority-day, eating-bird, Rule-activity, Adhikara, or Bharana products.

No object may silently acquire data from another profile. A source locator is part of computational truth, not decorative documentation. Admission status, capabilities, decision identity, and no-default policy belong to the manifest; normalized source facts remain in the profile document.

4. The 1879 Source-Scoped Profile

The governing source is the Tamil print catalogued by Internet Archive as dli.rmrl.000451_images, published in Madras in 1879 and traditionally attributed to Agastya. The traditional attribution is bibliographic metadata, not a verified historical authorship claim.

The admitted reading is deliberately narrow:

• the opening ontology is aksara/prasna: one explicitly listed initial vowel maps to a bird;
• Purva-day, Purva-night, Amara-day, and Amara-night remain four distinct regimes;
• machine-assisted visual reading directly maps waxing lunar phase to Purva at IA leaf n16 and waning phase to Amara at n26; that normalized mapping is source-scoped, its reading uncertainty remains explicit, and it carries no human-review dependency;
• each fixed half contains five samams of six nazhigai, totalling thirty nazhigai;
• activity durations are exact: Eat 5/4, Walk 3/2, Rule 2, Sleep 3/4, and Die 1/2 nazhigai;
• the schedule follows the machine-reconciled first-bird and assembly rules, including the corrected Purva-night assignment;
• identified printed-grid axes govern bird/activity assignments, while explicit prose and verse govern chronology; visual grid order is never chronological authority;
• the relationship product contains all twenty ordered non-self pairs, stores direction explicitly, and never infers reciprocity; and
• every generated cell retains governing temporal, assembly, chronology, and duration locators.

Stage 2A modern local-solar composition

The additive local_solar_day_explicit_paksha_v1 policy is Moira-owned modern composition policy, not a translated rule from the 1879 witness. Its governing object is PanchaPakshiLocalSolarContext with an embedded immutable PanchaPakshiLocalSolarContextPolicy.

The low-level engine receives an explicit profile, UT1 instant, latitude, longitude, and caller-supplied Purva or Amara label. Facade and REST callers instead provide a timezone-aware datetime that is normalized to UTC and crosses the established UTC civil-anchor boundary to UT1 once. The shared local-solar-day resolver then:

1. resolves configured-reader-backed topocentric sunrise, sunset, and next sunrise for the fixed sea-level observer elevation (0 m) using an unrefracted altitude signal and the existing -0.833-degree threshold that incorporates conventional standard refraction and solar semidiameter;
2. classifies [sunrise, sunset) as day and [sunset, next sunrise) as night;
3. derives the weekday from local mean solar time at the governing sunrise; and
4. selects the unchanged nominal source schedule from the explicit paksha plus derived half and weekday.

Polar geometry without lawful solar bounds fails explicitly. The context result exposes its UT1 boundaries and policy. Its provenance routing status is local_solar_half_and_weekday_performed_paksha_caller_supplied. It does not infer the paksha, scale source durations, convert nominal offsets to event instants, or identify a current schedule cell.

Stage 2B modern fixed-clock materialization

The additive fixed_24_minute_nazhigai_from_local_solar_half_start_v1 policy is also Moira-owned modern composition policy. It does not rewrite the unchanged 1879 profile or attribute its clock and astronomical assembly to that witness. Its governing object is PanchaPakshiFixedClockMaterialization with immutable PanchaPakshiFixedClockMaterializationPolicy and PanchaPakshiFixedClockCell members.

The policy composes four separately owned layers:

1. the named 1879 profile supplies the exact nominal thirty-nazhigai schedule, rational cell offsets, and source locators;
2. the University of Madras Tamil Lexicon, page 2231 defines one nazhigai as sixty vinadi or twenty-four minutes;
3. the Stage 2A local-solar context supplies the containing half, weekday, selected nominal schedule, and astronomical half boundaries; and
4. the IERS TT convention and IERS Technical Note 29 govern the TT/SI-second time-scale basis.

Day materialization anchors at the governing topocentric sunrise; night materialization anchors at the governing topocentric sunset. Every exact nominal offset is multiplied by 1,440 SI seconds/nazhigai, added on the configured reader’s TT path, and projected back to UT1 for publication. Thus thirty nazhigai always spans exactly 43,200 SI seconds on TT. Materialized cell intervals are half-open.

The fixed span is never clipped or stretched to the astronomical sunset or next sunrise. The result instead exposes fixed_end_jd_tt - solar_end_jd_tt as a signed-second residual and classifies it as before, coalescent, or after; absolute residuals no greater than 0.0001 s are coalescent under numerical policy. This topology is necessary because a twelve-hour fixed span can end before or after a seasonally unequal solar half. It is not an accuracy tolerance or historical assertion.

The result does not identify a current cell. The requested instant can remain inside the governing astronomical half while lying outside the materialized fixed span. No precedence is invented for that condition, and no solar-proportional scaling is performed or admitted by this fixed-clock result. The provenance routing status is fixed_clock_materialization_performed_paksha_caller_supplied_no_current_cell.

Stage 2C modern fixed-clock current-cell selection

The additive fixed_clock_current_cell_half_open_solar_precedence_v1 policy performs one bounded operation over the admitted Stage 2B materialization. Its governing object is PanchaPakshiFixedClockCurrentCellSelection, with immutable PanchaPakshiFixedClockCurrentCellSelectionPolicy and the finite PanchaPakshiCurrentCellSelectionStatus values selected and unmaterialized_solar_half_tail.

Selection order is doctrine, not an implementation accident. Stage 2A first resolves the half-open governing local-solar half. Stage 2B then materializes that half’s fixed schedule. Only then is the requested instant converted once to reader-bound TT and compared with the materialized TT cells. A cell owns the instant exactly when start_jd_tt <= requested_jd_tt < end_jd_tt; shared endpoints belong to the following cell and the final fixed endpoint is excluded. Membership tolerance is exactly 0.0 s. The Stage 2B 0.0001 s topology coalescence remains a classification policy and never changes current-cell ownership.

Solar-half precedence resolves seasonal overlap honestly. At exact sunset or sunrise, the newly governing half is selected before cell membership is evaluated. Cells from a prior short solar half that extend beyond its solar end remain inspectable in that prior materialization but are ineligible to become current. Conversely, when a long solar half continues after the fixed span, the interval from the excluded fixed end to the excluded solar-half end has status unmaterialized_solar_half_tail and current_cell=None. The selector does not clip, wrap, repeat, stretch, borrow, or retain a cell to conceal that tail.

Paksha remains caller supplied. Solar-proportional scaling and astronomical paksha inference remain not_performed. The provenance routing status is fixed_clock_current_cell_selection_performed_paksha_caller_supplied_no_scaling_or_inference. This is a modern deterministic interval-membership claim, not an assertion that the 1879 witness specified Moira’s astronomical or time-scale composition.

Stage 2D modern solar-proportional materialization

The additive solar_proportional_nominal_offsets_over_governing_half_tt_v1 policy is a separate Moira-owned composition. Its governing object is PanchaPakshiSolarProportionalMaterialization, with immutable PanchaPakshiSolarProportionalMaterializationPolicy and PanchaPakshiSolarProportionalCell members. It does not change the fixed-clock Stage 2B or Stage 2C products.

Stage 2A supplies the governing half, astronomical bounds, weekday, and unchanged nominal schedule. For every exact nominal endpoint n in the thirty-nazhigai span N, Stage 2D retains the reduced fraction f = n / N. The governing anchor and solar-half end are converted through the same reader to TT. Every interior endpoint is computed independently as anchor_jd_tt + float(f) * (solar_end_jd_tt - anchor_jd_tt), rather than by accumulating cell durations, and is then projected to UT1. The zero endpoint and final endpoint are set exactly to the governing TT and UT1 bounds. All 25 cells are positive, contiguous, and half-open; the solar-half end is excluded.

The fixed 1,440 s nazhigai conversion is not used. Because the normalized fractions close on the complete governing half, no clipping, wrapping, repetition, borrowing, or tail fabrication occurs. Paksha remains caller supplied. Current-cell selection and astronomical paksha inference remain not_performed. The provenance routing status is solar_proportional_materialization_performed_paksha_caller_supplied_no_current_cell_or_inference.

The 1879 witness governs the nominal schedule and exact rational offsets, not the proportional sunrise-to-sunset mapping. Route-specific provenance therefore replaces the raw profile omission seasonal_scaling with source_attested_solar_proportional_materialization: the result states both that modern materialization was performed and that the source did not attest that policy, without making contradictory claims or rewriting earlier results.

The source reading is machine-assisted and has not been collated against a genuinely independent witness. Every public result therefore carries the profile’s source-scoped status, decision identity, source identity, assembly policy, astronomical routing status, and declared omissions. Human-language review is not an admission dependency.

Stage 2E modern solar-proportional current-cell selection

The additive solar_proportional_current_cell_half_open_solar_precedence_v1 policy is a separate Moira-owned selector over the unchanged Stage 2D materialization. Its governing object is PanchaPakshiSolarProportionalCurrentCellSelection, with an immutable PanchaPakshiSolarProportionalCurrentCellSelectionPolicy. It does not change Stage 2D materialization or either fixed-clock product.

Stage 2A resolves the governing solar half before selection. Stage 2D then materializes all 25 proportional cells for that half with the same configured reader. The requested UT1 instant is converted once to reader-bound TT and is tested using exact start_jd_tt <= requested_jd_tt < end_jd_tt membership with 0.0 s tolerance. The anchor belongs to cell zero, every shared endpoint belongs to the following cell, and the old half’s final endpoint is excluded. Exact sunrise or sunset is therefore resolved into the new governing half and selects its first proportional cell.

Complete Stage 2D coverage makes selected the only lawful status and the selected cell is always non-null. Zero or multiple matches signal corrupted computation and fail closed. There is no unmaterialized tail, clipping, wrapping, repetition, borrowing, fixed-clock fallback, tolerance, or cell retention. Paksha remains caller supplied; astronomical paksha inference and natal identity remain separate and unperformed.

Stage 2E provenance must equal the Stage 2D provenance with only its astronomical routing status replaced by solar_proportional_current_cell_selection_performed_paksha_caller_supplied_no_fixed_clock_mixing_or_inference. The Stage 2D omission source_attested_solar_proportional_materialization remains intact, because neither selection nor API admission retroactively makes the proportional rule source-attested.

Stage 2F source-mapped astronomical paksha inference

Stage 2F adds PanchaPakshiAstronomicalPakshaInference and its immutable PanchaPakshiAstronomicalPakshaInferencePolicy. Machine-assisted visual reading of the exact 1879 witness directly maps waxing to Purva at IA leaf n16 and waning to Amara at n26. This mapping belongs only to the named profile, keeps its machine-assisted reading uncertainty visible, and carries no human-review dependency; it is not independent-witness corroboration or a universal Pancha Pakshi vocabulary.

The numeric classifier is the modern Moira policy apparent_geocentric_moon_sun_longitude_paksha_half_open_v1. One UT1 instant is converted once to reader-bound TT, and apparent geocentric Sun and Moon longitudes are evaluated in the true ecliptic of date on that shared TT. Normalized Moon - Sun elongation assigns [0, 180) to Shukla/waxing/Purva and [180, 360) to Krishna/waning/Amara. Exact 0 belongs to Shukla/Purva and exact 180 to Krishna/Amara; tolerance and snapping are both absent. A common ayanamsa is not applied because it cancels from the longitude difference.

This product accepts no location or caller-supplied paksha. It returns only the instantaneous astronomical half, source-mapped profile label, direct locator, policy, UT1/TT and longitude witnesses, and provenance. It performs no schedule selection, materialization, current-cell selection, automatic routing into another operation, or natal identity.

Stage 2G source table and modern natal-Moon composition

Stage 2G adds PanchaPakshiNakshatraBirdMapping as a pure source-table vessel and PanchaPakshiNatalMoonIdentity with immutable PanchaPakshiNatalMoonIdentityPolicy as a separate composition. Rendered-page inspection of the Bogamuni-attributed 2024 Internet Archive original PDF found the complete Purva partition at leaf n52, the complete Amara verse at n64, and the source phase-to-Purva/Amara binding at n167. These mappings belong only to bogamuni_chennai_2024_nakshatra_natal_identity.

The commentary adjacent to the Amara verse is malformed: it duplicates Shravana and omits Revati. The profile declares verse_precedence_for_nakshatra_partition, admits the complete verse, and retains the commentary as rejected conflict evidence. This is text-layer precedence, not inferred symmetry or a repaired commentary. The Uromarisi-attributed 1934 witness corroborates the Purva grouping at leaf n18 and independently exhibits malformed Amara commentary at n61; it remains non-executable evidence and does not supply runtime cells.

The sources attest nakshatra-bird associations and phase labels, but they do not explicitly govern a birth-Moon computation, Lahiri ayanamsa, or the equal 27-sector numerical taxonomy. The policy therefore fixes and exposes composition_status="modern_moira_policy_not_source_claim". One explicit UT1 instant is converted once to reader-bound TT. Apparent geocentric Sun and Moon longitudes in the true ecliptic of date determine the half-open Shukla/Krishna phase and source Paksha. The same TT epoch governs Lahiri true ayanamsa and the sidereal Moon. Twenty-seven equal half-open 40/3-degree sectors assign exact internal boundaries to the following nakshatra. A maximum-one-ULP-below recovery is permitted only when binary representation places an exact mathematical boundary immediately below itself; it is not a tolerance band.

The result exposes every astronomical and sidereal intermediate, the phase locator, nakshatra placement, nested source-table bird mapping and locator, policy, provenance, and declared omissions. It performs no schedule selection, materialization, current-cell selection, score, or forecast. The 1879 astronomical-paksha route remains standalone and does not acquire natal semantics.

Stage 2H source-owned Padu-bird table

Stage 2H adds PanchaPakshiPaduBirdMapping and the separate bogamuni_chennai_2024_padu_bird_mapping profile. Rendered-page inspection of the exact Bogamuni 2024 Internet Archive original PDF, locally identified by SHA-256 035eab41f62cf078180c03e99ec9eacf8edf2d2dc6d3dc31b37e6a6dfdb09990, found the Purva weekday Padu stanza and commentary at leaf n52, the Amara material at n60, the repeated combined table at n157, and its restating commentary at n158. The declared assembly policy is paksha_stanzas_govern_repeated_combined_table_confirms: the Paksha stanzas govern, while the later combined table and commentary confirm rather than silently replace them.

The complete source-owned mapping is:

Paksha	Sunday	Monday	Tuesday	Wednesday	Thursday	Friday	Saturday
Purva	Owl	Crow	Cock	Peacock	Vulture	Owl	Vulture
Amara	Crow	Owl	Vulture	Peacock	Cock	Peacock	Cock

These are exactly fourteen (profile_paksha, weekday) cells. There is no day/night axis. The result calls the selected value Padu and retains the source semantics profile_paksha_weekday_death_or_inoperative_bird_not_schedule_rule_activity. It does not inspect or transform a nominal schedule, derive a weekday from an instant, select a current cell, or reinterpret instantaneous RULE activity.

The source layers also distinguish an eating-bird table and authority days; they do not state that either object is an Adhikara Pakshi table. Bharana is secondary-only terminology rather than the governing primary table label. Moira therefore admits no Adhikara/Bharana alias or product and does not relabel the pre-existing schedule field first_eat_bird. Uromarisi 1934 and the separately inspected Bogar material are unbound research context only: neither the Stage 2H profile nor its decision binds them, and they do not contribute runtime cells, create synonymy, or prove admission.

The product is a pure explicit-label lookup. It performs no astronomical or civil-day routing, natal identity, schedule/materialization/current-cell operation, condition evaluation, score, or forecast. It has no default and does not add Padu semantics to either earlier profile.

Stage 2I source-owned first-samam EAT seed

Stage 2I adds PanchaPakshiFirstEatBirdMapping and the granular first_eat_bird_mapping capability to the unchanged 1879 schedule profile. The computational object is exactly (profile_id, profile_paksha, half, weekday) -> first_eat_bird. Rendered-page inspection binds the four governing source leaves: n16 for Purva day, n21 for Purva night, n26 for Amara day, and n31 for Amara night. The schedule grids and continuation leaves already bound to each named generator are same-witness confirmation, not independent evidence.

Context	Sunday	Monday	Tuesday	Wednesday	Thursday	Friday	Saturday
Purva day	Vulture	Owl	Vulture	Owl	Crow	Cock	Peacock
Purva night	Crow	Cock	Crow	Cock	Peacock	Vulture	Owl
Amara day	Cock	Peacock	Cock	Crow	Owl	Vulture	Peacock
Amara night	Vulture	Cock	Vulture	Owl	Crow	Peacock	Cock

These are exactly 28 first-samam seeds. The bird begins the selected generator’s first samam in EAT; later samams advance under that generator’s separate progression rule. The result therefore does not describe an ambient or whole-day eating bird. It exposes the generator ID, exact input axes, first_eat_bird, canonical generator locators, and profile provenance without materializing the 25-cell schedule.

The Uromarisi 1934 publication is a physically separate artifact and corroborates all 28 cells at leaves n6 and n36-n37. Its textual-lineage independence from the related teaching tradition is not established, so it is separate-publication corroboration rather than runtime data, a universal-canon proof, or grounds for corroborated_public status. The earlier blind reading’s Amara-night vector is explicitly superseded by the later n31-n35 adjudication; the canonical profile contains the corrected vector above.

The source phrase adhikara nalil does not supply a separate authority-bird table or prove synonymy. Stage 2I therefore admits no Padu, authority, Adhikara/Bharana, condition, strength, scoring, electional, or forecasting semantics and composes none of the four profiles.

Stage 2J research-only vinadi recovery

Stage 2J records the source-owned object (activity, explicit vinadi ordinal 1..5, named context) from rendered review of Uromarisi-attributed 1922, 1932, and 1934 publications. These witnesses preserve governing verses 228 and 229 and organize contextual outcomes by first through fifth vinadi positions beneath EAT, WALK, RULE, SLEEP, and DIE. The 1922 Karunanidhi Pillai commentary reiterates that the nazhigai must be divided accurately but supplies no arithmetic selector. The 1932 contents also expose distinct birth/natal, illness, journey, and house-foundation context families. Separate publication does not establish independent textual lineage.

Rendered review of the Bogamuni-attributed 2024 editorial witness recovers two separately named and incompatible temporal selectors for one six-nazhigai samam:

• provisional policy bogamuni_2024_weighted_sookshma_samam_v1 uses activity durations EAT 3/2, WALK 5/4, RULE 2, SLEEP 3/4, and DIE 1/2 nazhigai, which close exactly to six, with cyclic order beginning at the named activity; and
• provisional policy bogamuni_2024_eka_sookshma_equal_fifths_v1 divides the samam’s nazhigai into five equal parts.

These policies are not synonyms and may not be blended or silently selected. Neither is automatically bound to the Uromarisi ordinal outcome tables; that would be a cross-witness composition requiring its own explicit admission. Machine-assisted reading also does not support normalizing or translating the context-specific prognostic statements. Stage 2J therefore remains research_only because selector choice and witness composition are unadmitted, not because no source-attested formula exists. There is no human-language-review dependency. Uromarisi clock routing, source-outcome normalization, automatic composition, and Uromarisi-linked public exposure continue to fail closed.

Stage 2K explicit Sookshma temporal-selector admission

Stage 2K converts only the two Bogamuni 2024 editorial selector candidates into one separate source-scoped public profile. The caller must select exactly one policy on every call:

• bogamuni_2024_weighted_sookshma_samam_v1 rotates EAT 3/2, WALK 5/4, RULE 2, SLEEP 3/4, and DIE 1/2 nazhigai from the named parent activity; and
• bogamuni_2024_eka_sookshma_equal_fifths_v1 divides the six-nazhigai samam into five exact 6/5-nazhigai ordinal cells without assigning activities.

elapsed_nazhigai is an exact reduced rational in [0, 6). Interval ownership is [start, end), so every lawful offset has exactly one match. The selected policy and all five intervals are returned. There is no default, automatic policy choice, clock or civil-time routing, astronomy, schedule composition, or Uromarisi-outcome binding. Outcome interpretation, condition, scoring, electional search, and forecasting remain unadmitted. Human-language review is not an admission or maintenance dependency.

Stage 2L independent-witness collation gate

Stage 2L is a completed research collation and a non-admission decision. It inspects the exact 574-page Sarasvati Mahal Library 2014 sixth edition of Valaiyarul Patinen Siddhargal Panchapakshi Sastram, the 89-page G. R. Narasimhan secondary guide supplied for review, and a second 21-page supplied guide that is rejected for unrelated tarot contamination. No artifact enters the package.

The Sarasvati editor states that the verses derive chiefly from library palm-leaf manuscripts and a minor share of printed books, while the explanations, mathematical notes, and tables include editorial contributions. The source inventory names Agastya manuscript 777, Kumbamuni manuscript 7015, and Bogar and Uromarisi material. The editor also explicitly records different activity-order doctrines between Agastya/Bogar and Uromarisi and selects Agastya/Bogar for most of the compilation. This makes the volume a strong multi-source conflict comparator, but not a textually independent witness: no stemma or copying history establishes independence from the already admitted witnesses.

The collation preserves both agreement and disagreement:

• all seven Purva-day first-EAT seeds agree with the 1879 profile;
• the waxing-day EAT/WALK/RULE/SLEEP/DIE vector agrees exactly at (5/4, 3/2, 2, 3/4, 1/2) nazhigai;
• the compilation and Narasimhan guide instead use waxing-night (5/4, 5/4, 1, 1, 3/2), waning-day (2, 3/2, 3/4, 1/2, 5/4), and waning-night (7/4, 7/4, 3/4, 3/4, 1) vectors; and
• the secondary guide supplies natal-star tables as well as aksara methods, which cannot reinterpret the admitted 1879 aksara/query product.

Publication separation, institutional custody, exact row agreement, and a modern guide’s repeated tables do not individually establish textual independence. The Stage 2L decision therefore remains research_only, leaves all four public profiles source_scoped_public, and changes no profile, manifest, capability, engine, facade, or REST surface. No human-language reviewer is required. A future corroboration attempt must obtain an accessible candidate such as the catalogued Ramadevar manuscript, establish its copying history, and collate one named product with repeatable locators before any separate admission decision.

Stage 2M Ramadevar candidate identity and access gate

Stage 2M resolves the catalog identity without inferring manuscript content. The Commissionerate of Indian Medicine and Homoeopathy catalog identifies serial 859, manuscript A5, as Ramadevar Panchapakshi on PDF page 52. The catalog provides no item-level leaf images, transcription, physical description, date, incipit, explicit, colophon, provenance chain, copying history, or computational rules. The candidate is therefore exactly identified but remains catalog_title_only; product comparability and textual independence are not_assessable.

Similar names must remain separate objects. G.O.M.L. R.8978 Ramadevar Patchini is catalogued as a gnana, breath, and yoga work. British Library EAP EAP1217/1/2851, Tamil University reference TU_TAMIL_2058-01_2661, is an eighteenth-century 108-poem work on philosophy and ashtanga yoga. The same Commissionerate catalog lists a different Ramadevar Patchani - 108 inside manuscript 27, and the 1991 Sarasvati Mahal Ramadevar Sutiram: Ashtanga Patchini, Valai Pujai governs medicine, alchemy, yoga, and ritual. None is the five-bird temporal product merely because it shares a Ramadevar attribution, Patchani wording, or a 108-poem count.

The required acquisition object is specifically Commissionerate manuscript A5: a complete facsimile or critical transcription plus item-level physical, scribal, custodial, and copying-history metadata. Only after one named Pancha Pakshi computational product can be collated cell by cell with repeatable leaf locators may a separate admission decision be considered. Stage 2M changes no profile, manifest, capability, engine, facade, or REST surface and creates no human-language-review dependency.

Stage 2N explicit schedule-to-Sookshma composition

Stage 2N is the separate composition decision required by the earlier no-blending rule. The caller names the admitted 1879 schedule profile, admitted Bogamuni 2024 selector profile, source Paksha, half, weekday, samam 1..5, subject bird, selector policy, and exact elapsed Fraction within the samam. Moira generates the schedule, requires exactly one cell for that subject bird in that samam, and uses the cell’s activity as the Stage 2K parent activity.

The composition policy is explicit_schedule_samam_subject_bird_sookshma_v1 and is explicitly modern_moira_policy_not_source_claim. Neither source is credited with the cross-profile join. Both source provenances remain present in the nested result. There is still no default, clock or civil-time routing, astronomical routing, Uromarisi outcome binding, translation, condition, scoring, electional search, or forecasting. The profile documents, capabilities, and manifest are unchanged. The decision fixture SHA-256 is 084190606dc358abce7cc1879aa898a0071bce421b1eda8845b113520a7c36a9.

Stage 2O explicit civil-time-to-Sookshma routing

Stage 2O is a separate modern routing decision layered on Stage 2N. The caller must name both profiles, aware civil instant, location, source Paksha, subject bird, timing policy, and selector policy. Timing is exactly one of the already admitted fixed-clock or solar-proportional materializations; neither is a default and no fallback is allowed.

The current materialized cell determines the samam. Moira lifts the stored binary64 requested, samam-start, and samam-end reader-bound TT values exactly to Fraction, computes (requested - start) / (end - start) * 6, and supplies that exact offset to Stage 2N. Shared boundaries remain half-open and therefore belong to the following samam. A fixed-clock long-half tail returns the existing unmaterialized_solar_half_tail with null samam, offset, and composition; it is never silently filled by the proportional route.

The routing policy is civil_time_materialized_samam_to_stage2n_v1 and is explicitly modern_moira_policy_not_source_claim. Paksha remains caller supplied. No astronomical paksha inference, Uromarisi outcome binding, translation, condition, scoring, electional search, or forecasting occurs. Profiles, capabilities, and manifest remain unchanged. The decision fixture SHA-256 is 2ea686e774ba4468c0515f621771b8a142c79f04d89b69839f482e05c37b40df.

Stage 2P research-only Uromarisi illness-grid recovery

Stage 2P selects one Stage 2J context family for structural recovery without admitting its outcome semantics. The primary witness is the exact 166-page 1934 Uromarisi-attributed Vinadi Pancha Pakshi Mulamum Uraiyum PDF, locally verified at SHA-256 e2ab7a64d4d4e540c30bc464c12923e6f14e93fbbe15d73e459e9c62a5815da0. Rendered review of PDF pages 115–126 established a provisional 25-locator illness grid. Stage 2T later refines verse 250: its heading and verse assign DIE ordinal 5 while its commentary assigns SLEEP ordinal 5, so it is a blocked identity-conflict candidate rather than an unconflicted SLEEP cell.

The 25 cell locators are verses 230–239 and 241–255. Verse 240 is an intervening transition rather than an ordinal cell heading. Verse 256 begins a separate illness-duration section after the grid. Each cell records only its activity, ordinal, verse, and PDF/printed-page span. Stage 2T preserves the verse 250 layer conflict without mutating the historical Stage 2P fixture. Tamil source expression, translations, and normalized outcome payloads are not distributed or admitted.

This grid does not supply selector arithmetic and does not establish that either Stage 2K Bogamuni policy belongs to the Uromarisi lineage. Stage 2O therefore cannot route into it. Outcome labels, condition mapping, numeric scores, medical advice, prognosis, electional judgment, and forecasting remain forbidden. The manifest and all public surfaces are unchanged. No human-language reviewer is an admission dependency. The research decision SHA-256 is 449efb11b81741e1ac591d6a93033023f67892ac835cbcb178103606eb729dd2.

Stage 2Q research-only EAT-cell semantic-atom pilot

Stage 2Q applies a bounded source-owned vocabulary to only the five EAT cells from the Stage 2P illness grid. The governing object is a cell-local record of what the historical witness states: resolution, stated duration, distinct prescribed-response categories, medicine and prithivi term references, unresolved relation clauses, and uncertainty. It is not a generic favorable or unfavorable label and is not a condition or numeric score.

Rendered PDF pages 116–118 control the reading. Archive.org OCR line spans are retained only as navigation aids. The source-stated durations are 4 or 5, 7, 9, 13, and 15 days for ordinals one through five. Each cell is machine-checked by its exact activity, ordinal, verse, page tuple, OCR navigation span, semantic atoms, and canonical fixture hash. Tamil source expression and full translations remain outside the distributed artifact.

This pilot describes a historical text; it neither validates its medical claims nor offers diagnosis, prognosis, advice, or treatment. Disputed relation clauses remain present with null semantics. No selector is bound, no Stage 2O result is routed to an outcome, and no runtime profile, data, capability, engine, facade, or REST surface changes. No human-language reviewer is required. The research decision SHA-256 is 7b4311912ece7f49b30773604c91537ca5fa2a9e02b75baeebfb5bdc2575bcd9.

Stage 2R research-only WALK-cell semantic-atom extension

Stage 2R applies the same cell-local method to WALK verses 235–239 while extending the ontology only where the text requires it. Ordinals one and four state resolution; ordinals two and three state abatement; ordinal five gives a timed progression without an explicit resolution statement. These dispositions must not be collapsed into one generic outcome.

The time expressions are exactly 10 days, 15 days, within 20 days, 25 days, and within one month. Exact and upper-bound forms remain distinct, and a month is never silently converted into days. Source deity titles and their stated relations, explicit response categories, medicine or physician references, water-clause roles, the Navagraha-dosha reference, and unresolved activity relations remain cell-local and uncertainty-bearing.

Rendered PDF pages 118–120 control the reading; Archive.org OCR lines are navigation aids only. This is a historical-text record, not medical validation, diagnosis, prognosis, advice, treatment, condition scoring, or forecasting. No selector, Stage 2O routing, runtime profile, manifest, capability, engine, facade, or REST surface changes. No human-language reviewer is required. The research decision SHA-256 is 361a0a334a73623cb0b2c1b0e73489db2c20d3c259e04540a303510113f0e0d6.

Stage 2S research-only RULE-cell semantic-atom extension

Stage 2S applies the cell-local method to RULE verses 241–245. All five cells state resolution, with time expressions of 3 days, 5 days, within 8 days, 10 days, and 12 days. The third cell remains an upper bound and must not be rewritten as an exact eighth-day event.

The source adds fire-clause roles, Saturn-dosha references, bounded historical effect language, and one surface no-enmity statement. These remain separate from source deity titles and explicit actions. No fire or dosha clause becomes a medical cause, no effect phrase becomes a symptom or score, and the relation statement is not bound to runtime doctrine.

Rendered PDF pages 120–122 control the reading; Archive.org OCR lines are navigation aids only. This is a historical-text record, not medical validation, diagnosis, prognosis, advice, treatment, condition scoring, or forecasting. No selector, Stage 2O routing, runtime profile, manifest, capability, engine, facade, or REST surface changes. No human-language reviewer is required. The research decision SHA-256 is 85142480188a00ddec3de6f192a36025f282ca0eefa4643a6f1d74da4cec811d.

Stage 2T research-only SLEEP semantic atoms and identity conflict

Stage 2T reviews SLEEP verses 246–250 but produces only four unconflicted semantic records. Verses 246–249 preserve resolution difficulty, recurrence, an exact 20-day expression, two upper-bound day expressions, and a conditional three-month mortality-or-resolution branch. Mortality, recurrence, distress, harm, wind-dosha, treatment, ritual, and activity-relation language remain historical atoms rather than prediction, diagnosis, cause, symptom, score, advice, or runtime doctrine.

Verse 250 is blocked by text-layer identity conflict. Its heading and verse assign DIE ordinal 5, while its commentary assigns SLEEP ordinal 5. No layer is silently preferred, and neither its mortality language nor its candidate five-day commentary expression is normalized into a cell payload. The Stage 2P fixture remains unchanged as historical locator evidence; current truth is four unconflicted SLEEP cells plus one blocked candidate.

Rendered PDF pages 122–124 control the reading; Archive.org OCR lines are navigation aids only. No selector, Stage 2O routing, runtime profile, manifest, capability, engine, facade, or REST surface changes. No human-language reviewer is required. The research decision SHA-256 is 09f7651325cdac058d9816b85b031ef528f514ae91fb8cd9636452b8d7fb302a.

Stage 2U research-only DIE semantic atoms

Stage 2U records five unconflicted DIE ordinals from verses 251–255. Verse 250 is not used as a sixth ordinal or silently repaired: it remains only the blocked Stage 2T precursor whose heading and verse say DIE ordinal 5 while its commentary says SLEEP ordinal 5.

The first DIE cell preserves a conditional mortality upper bound of two months and a separate six-month source tanmai branch whose meaning is not normalized. The second preserves both an instantaneous marker and a one-year marker without harmonizing them. The other three cells contain no numeric time expression. Mortality, life-departure, body-destruction, nonresolution, space/void, fate, deity, activity-relation, and unclear source-branch clauses are bounded historical atoms, not prediction, diagnosis, prognosis, cause, symptom, score, advice, deterministic fate, or runtime doctrine.

Rendered PDF pages 124–126 control the reading; Archive.org OCR lines are navigation aids only. Verse 256 begins a separate illness-duration section and terminates Stage 2U scope. No selector, Stage 2O routing, runtime profile, manifest, capability, engine, facade, or REST surface changes. No human-language reviewer is required. The research decision SHA-256 is 4954c13c33aa755bc0e8c6f47b7825d6ddb8346a2b6df39edf804872e81cbf70.

Constitutional Phase 2 classification closure

Phase 2 closes at the private research boundary over the 24 unconflicted illness-context records. The data-free internal module moira._pancha_pakshi_classification owns the typed descriptive model:

• PanchaPakshiHistoricalDisposition preserves exact source-statement classes rather than favorable/unfavorable labels;
• PanchaPakshiHistoricalTimeClass distinguishes exact, upper-bound, conditional, multiple, unreconciled, and unstated time shapes without selecting a temporal policy;
• PanchaPakshiHistoricalSemanticMarker records only the presence of named response, mediation, elemental/dosha, deity/fate, effect, activity-relation, mortality, and source-branch atoms;
• PanchaPakshiHistoricalCellClassification binds those classes to one activity, ordinal, verse, source decision/hash, and nonempty uncertainty;
• PanchaPakshiHistoricalIdentityConflict carries disagreeing text-layer identities without a classification payload; and
• PanchaPakshiUromarisiPhase2ClassificationCorpus enforces exact 5/5/5/4/5 coverage, uniqueness, deterministic order, and the blocked verse 250 identity.

Every closure row is machine-derived from the exact Stage 2Q–2U semantic atom: EAT uses its resolution and duration fields, WALK/RULE/SLEEP use their disposition and time-expression fields, and DIE uses mortality form and time-expression fields. Presence markers and uncertainty counts are likewise projected from the source record. Construction fails on inconsistent activity/disposition, mortality marker, uncertainty, coverage, order, identity, or conflict state.

This closure opens Phase 3 only for private derived convenience views and vessel consistency/malformed-construction hardening. Phase 3 may not add doctrine, interpretation, generic judgment, condition, score, prognosis, medical advice, temporal policy, Uromarisi selector binding, runtime profile, public export, facade method, or REST route. Selector doctrine belongs to constitutional Phase 4. Human Tamil review is not required; Commissionerate A5 acquisition gates only a future corroborated_public claim; and verse 256 begins a separately scoped product. The Phase 2 closure SHA-256 is a5cd64696d4c040554f2c235056dfd28477fd0796fc82306f44ae43473d434e2.

Constitutional Phase 3 inspectability closure

Phase 3 consumes only the frozen private Phase 2 vessels. It adds no source record or classification. Its cell properties expose:

• (activity, ordinal) identity and (decision ID, SHA-256) source binding;
• deterministic lexical semantic-marker values;
• exact stored presence of mortality language; and
• whether the existing time class is stated, conditional, or explicitly unreconciled.

Conflict properties expose heading, verse, and commentary assignments, their distinct activities, heading/verse agreement, and source binding without selecting a governing layer. Corpus properties expose classified and blocked verses, activity counts, unique source bindings, mortality-language cells, and unstated-time cells. Typed lookups return cells by activity or identity and return classifications and conflicts by verse. Absence returns None; no adjacent, symmetric, or cross-activity fallback exists.

Phase 3 hardens immutable tuple containers, exact EAT 230–234, WALK 235–239, RULE 241–245, SLEEP 246–249, and DIE 251–255 identity matrices, activity-specific Phase 2 time classes, one source binding per activity, and the Stage 2T source binding of the verse 250 conflict. Thus classification_at(SLEEP, 5) and classification_for_verse(250) are absent, while conflict_for_verse(250) remains present.

These are derived views and consistency rules only. They are not prediction, prognosis, condition, score, advice, outcome selection, temporal policy, or runtime admission. No package, facade, REST, manifest, or profile surface changes. Phase 3 completion permits Phase 4 to make one explicit doctrine or policy decision, but supplies no default or inherited Uromarisi selector binding. The Phase 3 decision SHA-256 is 2fd93585f8d2d439882ee77cdeb28e5509e916cd752357d60caaa003cc9fb2ca.

Constitutional Phase 4 doctrine and policy closure

Phase 4 admits one private research policy: moira_explicit_uromarisi_activity_ordinal_lookup_research_v1. The typed PanchaPakshiHistoricalClassificationPolicy requires PanchaPakshiHistoricalClassificationPolicyId.EXPLICIT_ACTIVITY_ORDINAL and exposes these fixed boundaries:

• the policy is a modern Moira research rule over source-owned ordinals;
• activity and ordinal are caller supplied explicitly;
• no temporal selector is source-attested or admitted;
• Stage 2K composition is not admitted;
• outcome interpretation is not performed;
• medical use is forbidden; and
• admission remains research_only.

pancha_pakshi_uromarisi_classification_under_policy requires corpus, policy, typed activity, and ordinal with no default. It delegates exactly to the Phase 3 activity/ordinal lookup and preserves None without fallback. It accepts no instant, clock, elapsed duration, samam offset, or selector policy. SLEEP ordinal 5 therefore remains absent and verse 250 remains only a conflict.

The Bogamuni 2024 weighted-Sookshma and Eka-Sookshma equal-fifths policy IDs remain visible as separate unadmitted cross-witness candidates. Neither is attributed to the Uromarisi witness, selected automatically, or routed into the historical classifications. A future composition would require its own explicit governing object, provenance, non-medical semantics, and admission decision.

Phase 4 adds no relation doctrine, condition, score, prognosis, advice, runtime profile, manifest entry, package export, facade method, or REST route. Its completion permits Phase 5 relational formalization only; no unresolved activity-relation clause is silently resolved. The Phase 4 decision SHA-256 is 4a444c91bab9a4949664e6bca4e64ad0ee341b439019db831429e4548bd2c4f9.

Constitutional Phase 5 relational formalization

Phase 5 creates a private typed PanchaPakshiHistoricalRelationRecord for each of the 24 unconflicted Phase 2 classifications and binds them in PanchaPakshiUromarisiPhase5RelationCorpus. A record is owned by the classified activity, ordinal, and verse and must retain that classification’s exact source decision identity and digest.

Relation presence projects only the Phase 2 activity_relation_clause marker. The complete corpus contains 17 present and 7 explicitly not-recorded clauses. Ten EAT/WALK records retain unresolved_clause plus their stated high or medium confidence. Seven RULE/SLEEP/DIE records preserve only an exact bounded surface category: no_enmity, rule_enmity_disallowed, rule_enmity_branch, earth_rule_enmity_disallowed, or rule_enmity_required. Because those bounded atoms state no relation-specific confidence, the typed record must use not_stated; it may not invent one.

Neither presence nor a surface category establishes relation endpoints, direction, normalized meaning, favorability, condition, score, prognosis, or runtime semantics. A not-recorded clause cannot be rewritten as present, an unresolved clause cannot be normalized, and verse 250 remains outside the relation corpus as an identity conflict. The corpus must match all Phase 2 cells in canonical order, preserve source bindings exactly, and fail closed on any marker, ordering, or binding mismatch.

Phase 5 adds no selector binding, runtime profile, manifest entry, package export, facade method, REST route, prediction, diagnosis, advice, condition, score, or medical-truth claim. Its completion permits Phase 6 relational hardening and inspectability only. The Phase 5 decision SHA-256 is e8e189f75418cc96bc6930e2e93d2cfcebc849cb4080001ee4b4b07b158908d1.

Constitutional Phase 6 relation hardening and inspectability

Phase 6 adds derived-only views to the Phase 5 vessels. A relation record may expose its exact activity/ordinal/verse identity, source binding, detection, admission, scoring, unresolved-clause status, and named-surface status. These properties must read stored Phase 5 state and must not reconstruct or normalize source semantics.

The corpus may expose deterministic record identities and source bindings; separate present and not-recorded counts; exact detected, not-recorded, unresolved, named-surface, admitted, and scored subsets; typed activity filtering; and exact activity/ordinal and verse lookup. Detected means only presence == present. It does not imply semantic admission. Admission does not imply scoring. At this boundary the detected subset contains 17 records while the admitted and scored subsets are both empty.

Missing lookup returns None without fallback. SLEEP ordinal 5 remains absent, verse 250 remains the blocked classification conflict, and neither may be repaired into a relation record. Invalid activity types, ordinals, verses, containers, and member types fail explicitly. Derived collections are immutable tuples and preserve canonical Phase 5 ordering.

Phase 6 adds no source truth, relation meaning, endpoint, direction, condition, score, selector binding, runtime profile, manifest entry, package export, facade method, REST route, prediction, diagnosis, advice, or medical-truth claim. Its completion permits Phase 7 integrated-local-condition work only; Phase 7 inherits no favorable/unfavorable or scoring doctrine. The Phase 6 decision SHA-256 is b175bcd1e537fb551cd26b18d6e6caa37f7a574b7e0a96b336d6fbb97eff9b12.

Constitutional Phase 7 integrated local condition

Phase 7 defines PanchaPakshiHistoricalLocalConditionProfile as one exact classification/relation pair under the explicit Phase 4 activity/ordinal policy. PanchaPakshiUromarisiPhase7LocalConditionCorpus contains one such profile for every unconflicted relation record. Materialization requires the typed policy and has no default or temporal-selector binding.

The classification and relation must share activity, ordinal, verse, source decision identity, and source digest. The corpus must preserve the canonical classification and relation order, retain the same policy on every profile, use an immutable tuple, reject non-profile members, and cover all 24 cells. SLEEP ordinal 5 and verse 250 remain absent without fallback or repair.

This local condition is a structural integration only. Its sole evaluation status is not_evaluable_no_admitted_condition_doctrine; favorability is not_assigned, condition score is None, prognosis is not performed, medical use is forbidden, and admission remains research-only. Relation detection, admission, and scoring are visible separately. Since Phase 6 admits and scores zero relations, Phase 7 rejects injected admitted or scored relation state.

Phase 7 adds no source truth, relation meaning, evaluative doctrine, generic good/bad label, numeric score, selector binding, runtime profile, manifest entry, package export, facade method, REST route, prediction, diagnosis, advice, or medical-truth claim. Its completion permits Phase 8 aggregate structural intelligence only. The Phase 7 decision SHA-256 is 401c90b7d7c15663427e034a527f983054800f948019cf12b404a0086b3203be.

Constitutional Phase 8 aggregate intelligence and sequence closure

Phase 8 defines PanchaPakshiUromarisiPhase8AggregateIntelligence as a frozen structural summary derived from the complete Phase 7 corpus. Its materializer accepts only PanchaPakshiUromarisiPhase7LocalConditionCorpus and counts stored facts without reclassifying, reinterpreting, ranking, or weighting profiles.

The exact aggregate is 24 profiles; activity coverage EAT 5, WALK 5, RULE 5, SLEEP 4, and DIE 5; evaluation status not_evaluable_no_admitted_condition_doctrine for all 24; relations 17 detected, 7 not recorded, 10 unresolved, 7 named-surface, 0 admitted, and 0 scored; and blocked verse tuple (250,).

Activity and evaluation counts must cover all profiles. Detected plus not-recorded must cover the corpus; unresolved plus named-surface must cover the detected subset; scored must remain a subset of admitted and detected; admitted and scored must both remain empty. Canonical activity order, repeatability, and immutability are mandatory. A count cannot become a rank, weight, judgment, score, prognosis, or source meaning.

Phase 8 adds no source truth, evaluative doctrine, selector binding, runtime profile, manifest entry, package export, facade method, REST route, prediction, diagnosis, advice, or medical-truth claim. It closes the agreed eight-phase private sequence. Phase 9 network work is not automatic and public exposure is not authorized; either requires a separate explicit decision. The Phase 8 decision SHA-256 is b193b3ba62d1c5eb57d526777310fa16f29c81fa999e69b813670c680ba2fd13.

Constitutional Phase 9 network intelligence

The explicitly authorized Phase 9 projection contains one immutable node for each of the 24 Phase 7 profiles and one candidate annotation for each of the 17 detected source clauses. Node identity and order preserve activity, ordinal, verse, and source binding exactly. A candidate remains owned by its source node and is not an edge.

Because endpoint and direction remain not_established, admitted and scored edge tuples are empty, topology is not materialized, and graph metrics are not_evaluable_no_admitted_relation_edges. Absence is not a zero-weight edge; surface categories do not become endpoints; and verse 250 remains blocked. The Phase 9 decision SHA-256 is 49935df6e96595b5cd365dbea12acabcf862eb81a120c3d9122d29ad4962872b.

Constitutional Phase 10 full-subsystem hardening

Phase 10 freezes the complete source-to-network receipt chain. It preserves policy identity, canonical node IDs, five source decision bindings, 24 profiles, 17 candidates, empty admitted/scored edges, and blocked verse 250. The canonical structural fingerprint is 2133ad1c72ea5209facbb83ff8f40cfd09c1efea5340df7943fe08ff599cface. Malformed types, reordered or duplicated members, forged source bindings, mismatched aggregates, invented edges, implicit defaults, conflict repairs, and numeric substitution for unavailable metrics fail closed. The Phase 10 decision SHA-256 is 9ef977585ad1dc9dc517316eb864a8de26f462fb852977bfef936d8756ef64a0.

Constitutional Phase 11 architecture freeze and validation codex

The formal backend architecture is frozen in PANCHA_PAKSHI_UROMARISI_BACKEND_STANDARD.md. It records executable terminology, ordering, policy, relation, condition, network, hardening, failure, validation, public/private, and nonclaim doctrine. The standard admits only immutable governance status as a Phase 12 candidate; historical and network data remain private. The Phase 11 decision SHA-256 is 697eecaf22cf4e8d42ca9b7044633e6407ca8d5577dd4407180029cfc00055c0.

Constitutional Phase 12 public API curation

Phase 12 exposes PanchaPakshiUromarisiConstitutionStatus and pancha_pakshi_uromarisi_constitution_status() through the package and facade, and exposes that same governance-only product through GET /v1/pancha-pakshi/constitution/uromarisi. The immutable result reports completion of SCP Phases 1-12, research_only admission, private historical/network data, unadmitted relation semantics, unavailable graph and condition evaluation, no prognosis, and forbidden medical use.

This is governance metadata, not a runtime profile. The strict REST response adds no manifest entry, research fixture I/O, historical cell, relation candidate, edge, score, or interpretation. The Phase 12 decision SHA-256 is 47ddf7348a235245ca89dc637a20be37b7bec244aa1ded968c9c04b884d05a6d.

5. Conflicts That Forbid Blending Or A Default

The following are different computational doctrines or unresolved editorial defects, not values that may be averaged or repaired by symmetry.

Area	Evidence	Required policy
Identity ontology	The 1879 opening uses query/name initials; later presentations use birth nakshatra and Moon paksha.	Never label the 1879 mapping as natal identity.
Amara nakshatra partition	The Bogamuni 2024 verse is complete while its adjacent commentary duplicates Shravana and omits Revati.	The named 2024 profile declares verse precedence and rejects the malformed commentary; never blend or repair the layers.
Duration vectors	The 1879 witness has one vector; later witnesses contain regime-specific or non-closing vectors.	Durations belong to the named witness and text layer.
Vinadi temporal selectors	The 2024 editorial witness separately names a weighted Sūkṣma selector and an equal-fifths Eka Sūkṣma selector; Uromarisi ordinal witnesses supply no arithmetic binding.	Preserve both named policies, select neither by default, and do not attach either to Uromarisi outcomes without an explicit cross-witness composition decision.
Ramadevar title identity	Catalogs separately identify Panchapakshi, Patchini, Patchani 108, and Ashtanga Patchini works under Ramadevar attribution.	Never merge by attribution, spelling similarity, or poem count; require exact manuscript identity and product-level content.
Schedule cells	A later edition contains a verse/commentary mismatch that duplicates one bird and omits another.	Fail the affected layer; do not infer the missing bird.
Relationships	The 1879 table is directed and nonuniform; later tables use different assignments.	No reciprocity, symmetry inference, or cross-witness merge.
Timing	The inspected 1879 source attests fixed thirty-nazhigai halves, not proportional sunrise-to-sunset scaling.	Seasonal scaling, if admitted, is a separate policy and capability.
Padu, first-EAT, and authority vocabulary	The Bogamuni weekday Padu table and the 1879 first-samam EAT seeds are separately complete; primary layers mention authority days without supplying an Adhikara Pakshi table, and Bharana is secondary-only terminology.	Preserve the Padu and first-samam EAT objects as separate source-scoped lookups. Do not manufacture an Adhikara/Bharana alias or derive a day ruler from either lookup or instantaneous Rule activity.
Vinadi ordinal routing	Uromarisi-attributed 1922, 1932, and 1934 publications attest first through fifth vinadi result positions beneath activities without arithmetic; the 2024 editorial comparator separately attests weighted and equal-fifths selectors.	Preserve the ordinal axis and both selector candidates. Stage 2N permits only its separately decided modern schedule composition; do not infer Uromarisi binding, choose a default, normalize prognostic prose, or attach outcomes.
Verse 250 parent activity	The heading and verse assign DIE ordinal 5 while the commentary assigns SLEEP ordinal 5.	Preserve all layers, block semantic normalization, and do not assign the candidate to either parent activity at runtime.

Conflict-ledger witnesses and rejected text layers are not executable merely because they are recorded. The named 2024 profile admits only its declared governing verse and source mappings; Uromarisi evidence and the rejected commentary do not silently broaden either public profile.

6. Fail-Closed Invariants

A loadable profile must satisfy all applicable invariants exactly:

• known profile and manifest schema versions and matching canonical SHA-256;
• finite admission-status, product-kind, and capability registries;
• a nonempty admission-decision ID and default_selection_allowed=false;
• capabilities exactly consistent with the declared product kind;
• explicit profile ID, witness, derivation state, and assembly policy;
• four complete paksha/day-night regimes for this operating-schedule product;
• exactly seven weekday first-bird assignments per regime;
• exactly five known birds and five known activities;
• five samams per half;
• every samam contains every bird and activity exactly once;
• every half assigns every bird/activity pair exactly once;
• exact positive reduced rational durations totalling six nazhigai per samam and thirty nazhigai per half;
• nonempty source locators for every generated cell;
• no unknown, duplicated, overlapping, or omitted identity symbol;
• exactly one stored relationship for every ordered non-self pair, with no reciprocal inference; and
• no source component from a different profile.

For the additive local-solar context capability, the profile must also admit astronomical_context; paksha must remain an explicit source label; the instant and coordinates must be finite and lawful; the solved bounds must satisfy sunrise < sunset < next sunrise and contain the requested instant; the derived half and weekday must select one existing nominal schedule; and no offset materialization may occur.

For the additive fixed-clock capability, the profile must also admit fixed_clock_materialization; the Stage 2A context and explicit paksha rules remain governing inputs; day must anchor at sunrise and night at sunset; one nazhigai must remain exactly 1,440 SI seconds on reader-bound TT; every published endpoint must be the corresponding TT endpoint projected to UT1; all 25 intervals must be contiguous and half-open; the fixed span must remain exactly 43,200 TT seconds; no interval may be clipped or proportionally scaled to the solar end; and the signed TT end residual must obey the declared 0.0001 s coalescence policy. The Stage 2B result may claim no current cell.

For the additive fixed-clock current-cell capability, the profile must also admit fixed_clock_current_cell_selection; the complete Stage 2B materialization and caller-supplied paksha remain governing inputs; the requested instant must be converted to reader-bound TT once; membership must use exact half-open TT comparisons with 0.0 s tolerance; and the governing solar half must be resolved before selection. A selected result must carry one cell from that materialization. An unmaterialized_solar_half_tail result must carry no cell and must lie from the excluded fixed end to the excluded end of a longer governing solar half. No prior-half cell, clipping, wrapping, repeating, stretching, proportional scaling, or paksha inference is permitted.

For the additive solar-proportional capability, the profile must also admit solar_proportional_materialization; the complete Stage 2A context, unchanged nominal schedule, and caller-supplied paksha remain governing inputs; every nominal endpoint fraction must remain exact, reduced, ordered, and bounded by zero and one; and every TT endpoint must be derived independently from the common anchor and complete governing-half TT span. The first and final TT and UT1 endpoints must equal the governing solar-half bounds exactly. All 25 cells must be positive, contiguous, and half-open. The fixed nazhigai-second rule, current-cell selection, clipping, wrapping, repetition, tail fabrication, and paksha inference are prohibited. Result provenance must distinguish performed modern composition from absent source attestation.

The detached public result can prove its exact fraction-to-TT mapping, outer UT1 closure, and UT1 ordering/contiguity from its own fields. It cannot replay the reader-dependent inverse for an interior TT endpoint after the configured reader has been detached. Interior TT-to-UT1 truth therefore belongs to the governing factory path and its reader-bound tests; it is not presented as a self-contained vessel invariant.

For the additive proportional current-cell capability, the profile must also admit solar_proportional_current_cell_selection; the complete Stage 2D materialization and caller-supplied paksha remain governing inputs; Stage 2A solar-half ownership must be resolved before selection; and the requested instant must be converted to reader-bound TT exactly once. Membership must use exact half-open comparisons with 0.0 s tolerance and produce exactly one matching member of the materialization. The only lawful status is selected and its current cell is non-null. Zero matches, overlaps, a foreign or copied cell, an unmaterialized-tail status, fixed-clock mixing, fallback, clipping, wrapping, repetition, borrowing, or paksha inference are errors.

For astronomical paksha inference, the profile must admit astronomical_paksha_inference and contain exactly the source-owned waxing/Purva/n16 and waning/Amara/n26 mappings. One finite UT1 instant must produce one reader-bound TT used by both geocentric body evaluations. Longitudes and normalized elongation must lie in [0, 360); exact half-open classification must agree with the astronomical enum, profile label, and its single locator. Location, tolerance, snapping, ambient ayanamsa, schedule selection, materialization, current-cell selection, automatic routing, and natal identity are prohibited.

For the Stage 2G profile, nakshatra_bird_mapping must contain exactly one entry for every (purva|amara, 0..26) pair, use canonical nakshatra names, and resolve each cell to exactly one bird and one governing locator. The complete Amara verse must govern while the malformed commentary remains rejected conflict evidence. Natal identity requires the matching capability, exact fixed policy, one reader-bound TT epoch, consistent tropical/ayanamsa/sidereal longitudes, mathematically coherent half-open phase and nakshatra ownership, one phase locator, and a nested mapping whose profile, Paksha, nakshatra, bird, and provenance equal the direct identity fields. Schedule, current-cell, scoring, and forecast claims are errors.

For the Stage 2H profile, padu_bird_mapping must contain exactly one entry for every (purva|amara, Sunday..Saturday) pair and no day/night key. Every result must equal the canonical profile cell, carry the three source locators for its Paksha stanza plus the repeated combined table and commentary, retain the declared stanza-precedence assembly policy, and report astronomical routing as not_performed. Schedule inspection, RULE-activity conversion, first_eat_bird relabelling, Adhikara/Bharana aliasing, natal identity, current-cell selection, scoring, and forecasting are errors.

For the Stage 2I capability, the unchanged 1879 profile must contain exactly one first-EAT seed for every (purva|amara, day|night, Sunday..Saturday) combination. Every result must equal the canonical generator’s weekday seed, carry that generator ID and its complete ordered source-locator tuple, and match both PanchaPakshiSchedule.first_eat_bird and the first samam’s EAT cell. Datetime or location routing, inferred Paksha, schedule materialization, Padu or authority aliasing, whole-day semantics, natal identity, condition, scoring, and forecasting are errors.

Unknown values, hash mismatches, incomplete tables, mixed profiles, capability drift, and unresolved verse/commentary conflicts are errors. They are not warnings or fallback opportunities.

7. Public Contract

The stable engine surface is moira.pancha_pakshi, with package-root and moira.vedic exports and Moira delegation. The REST family is /v1/pancha-pakshi.

The five Phase 1 operations remain kernel-free. Stage 2A adds the kernel-backed pancha_pakshi_local_solar_context_at(...), Moira.pancha_pakshi_local_solar_context(...), and POST /v1/pancha-pakshi/context/local-solar surfaces. The low-level function accepts UT1 JD; facade and REST inputs use a timezone-aware datetime normalized to UTC and expose the resolved UT1 values in the result.

Stage 2B adds the separately governed kernel-backed pancha_pakshi_fixed_clock_materialization_at(...), Moira.pancha_pakshi_fixed_clock_materialization(...), and POST /v1/pancha-pakshi/schedule/fixed-clock surfaces. This operation accepts the same explicit profile, aware datetime or UT1 instant, location, and caller-supplied paksha boundary as Stage 2A. It materializes the selected source-owned nominal schedule only under fixed_24_minute_nazhigai_from_local_solar_half_start_v1; no ambient or alternate timing policy is accepted.

Stage 2C adds the separately governed kernel-backed pancha_pakshi_fixed_clock_current_cell_at(...), Moira.pancha_pakshi_fixed_clock_current_cell(...), and POST /v1/pancha-pakshi/schedule/fixed-clock/current-cell surfaces. It accepts the same explicit profile, aware datetime or UT1 instant, location, and caller-supplied paksha, and only the fixed_clock_current_cell_half_open_solar_precedence_v1 policy. The result contains the complete Stage 2B materialization, requested TT witness, finite selection status, selected materialized cell or explicit null, and provenance.

Stage 2D adds the separately governed kernel-backed pancha_pakshi_solar_proportional_materialization_at(...), Moira.pancha_pakshi_solar_proportional_materialization(...), and POST /v1/pancha-pakshi/schedule/solar-proportional surfaces. They accept the same explicit profile, aware datetime or UT1 instant, location, and caller-supplied paksha. The result contains the complete Stage 2A context, policy, governing TT/UT1 anchor and end, TT half duration, 25 proportionally materialized cells with exact nominal fractions, and route-specific provenance. It does not select a current cell.

Stage 2E adds the separately governed kernel-backed pancha_pakshi_solar_proportional_current_cell_at(...), Moira.pancha_pakshi_solar_proportional_current_cell(...), and POST /v1/pancha-pakshi/schedule/solar-proportional/current-cell surfaces. They accept the same explicit profile, aware datetime or UT1 instant, location, and caller-supplied paksha, plus only the named Stage 2E policy. The engine result retains its complete governing Stage 2D materialization and returns the requested TT witness, immutable selection policy, one non-null selected cell, and transformed provenance. The compact REST result publishes the governing bounds and selected cell without duplicating all 25 cells.

Stage 2F adds the separately governed kernel-backed pancha_pakshi_astronomical_paksha_at(...), Moira.pancha_pakshi_astronomical_paksha(...), and POST /v1/pancha-pakshi/context/astronomical-paksha surfaces. The engine takes an explicit profile and UT1 JD; facade and REST take an explicit profile and aware datetime. The REST request also requires the exact Stage 2F policy ID. No form accepts a location or caller-supplied paksha, and no inferred label is passed to a schedule operation.

Stage 2G adds the kernel-free pure-table pancha_pakshi_nakshatra_bird_mapping(...) and Moira.pancha_pakshi_nakshatra_bird_mapping(...) surfaces, plus POST /v1/pancha-pakshi/mappings/nakshatra-bird, and the kernel-backed pancha_pakshi_natal_moon_identity_at(...), Moira.pancha_pakshi_natal_moon_identity(...), and POST /v1/pancha-pakshi/identity/natal-moon surfaces. The natal engine takes an explicit profile and UT1 JD; facade and natal REST take an explicit profile and aware datetime. Natal REST additionally requires the exact Stage 2G policy ID. The pure-table REST route instead requires explicit source Paksha and a zero-based nakshatra index. No form accepts location, caller-supplied bird/ayanamsa, schedule/current-cell controls, scoring, or forecast policy.

Stage 2H adds the kernel-free pure-table pancha_pakshi_padu_bird_mapping(...), Moira.pancha_pakshi_padu_bird_mapping(...), and POST /v1/pancha-pakshi/roles/padu surfaces. Each accepts an explicit profile, profile_paksha, and weekday only. No form accepts an instant, location, day/night half, schedule, activity, natal identity, Adhikara/Bharana alias, condition, score, or forecast input.

Stage 2I adds the kernel-free pure-table pancha_pakshi_first_eat_bird_mapping(...), Moira.pancha_pakshi_first_eat_bird_mapping(...), and POST /v1/pancha-pakshi/schedule/first-eat-bird surfaces. Each accepts only an explicit profile, profile_paksha, day/night half, and weekday. No form accepts an instant, location, inferred Paksha, Padu or authority role, schedule materialization, natal identity, condition, score, or forecast input.

All computations require profile_id. Public results are immutable and carry profile-owned provenance and omissions. Exact nazhigai values remain rational in the engine and serialize as {numerator, denominator} at the transport boundary. The current family contains four profiles, nineteen Moira methods (eleven kernel-free and eight kernel-backed), and nineteen REST routes.

Stage 2F accepts a datetime but no location and performs only instantaneous astronomical-paksha inference. The five location-bearing Stage 2A-E operations all require explicit paksha. No operation accepts a caller-supplied natal-Moon longitude, nakshatra, score, inferred name, sunrise, or timezone policy as an override. Stage 2G alone computes natal-Moon identity from the supplied instant under its fixed policy. The context operation returns only the selected nominal schedule; the fixed-clock operation returns all materialized half-open cell intervals and their solar-half topology without a current-activity judgment; and the current-cell operation returns only the unique cell under its named fixed-clock policy or the explicit unmaterialized tail status. The proportional materialization operation separately maps exact nominal fractions across the governing half, while its separately named selector returns the unique current proportional cell. Neither proportional operation borrows fixed-clock semantics. Stage 2H remains separate from every temporal product: it performs one pure Padu-table lookup and never routes that bird into a schedule, identity, condition, or forecast. Stage 2I likewise returns only one named generator’s first-samam EAT seed and never materializes the schedule or promotes that seed into a whole-day or authority role.

8. Evidence And Validation

The frozen blind reading, representative-grid reading, page-image adjudication, and machine reconciliation remain historical evidence. Public admission does not rewrite their earlier private status. The additive public admission decision links the former and current profile hashes, names the exact public claim and nonclaims, binds the evidence records, and records a computational projection over 10 identity symbols, 28 schedules, 700 cells, and 20 directed relationships. Admission and provenance metadata are intentionally excluded because the migration changes them.

The additive Stage 2A admission decision binds the unchanged profile hash and the frozen Phase 1 decision, records the manifest-only capability transition, and names the modern composition policy, assembly doctrine, authority boundary, and public nonclaims. It does not rewrite the Phase 1 fixture or recast modern astronomical policy as historical source evidence.

The chained Stage 2B admission decision binds the unchanged profile hash, the frozen Stage 2A decision and manifest, and the manifest-only addition of fixed_clock_materialization. It records the exact fixed-clock policy, the 1879 nominal-offset source leaves, the University of Madras Tamil Lexicon nazhigai definition, the IERS TT/SI-second convention, and the Stage 2A solar-anchor boundary without claiming that their composition is an 1879 rule.

The chained Stage 2C admission decision binds the unchanged profile hash, frozen Stage 2B decision and manifest, and the manifest-only addition of fixed_clock_current_cell_selection. It records the exact solar-half-first, half-open TT membership policy and the explicit unmaterialized-tail result without presenting deterministic selection as an external Pancha Pakshi oracle or independent-witness corroboration.

The chained Stage 2D admission decision binds the unchanged profile hash, frozen Stage 2C decision and manifest, and manifest-only addition of solar_proportional_materialization. It records the exact fraction-over-governing-half TT policy, independent endpoint derivation, outer closure, and route-specific omission resolution without attributing the modern composition to the 1879 witness.

The chained Stage 2E admission decision binds the unchanged profile hash, frozen Stage 2D decision and manifest, and manifest-only addition of solar_proportional_current_cell_selection. It records exact solar-half-first, half-open TT membership, selected-only/non-null semantics, and unchanged Stage 2D omission provenance without presenting the selector as source-attested doctrine or external current-cell parity.

The chained Stage 2F admission decision binds the frozen Stage 2E decision and prior manifest, the source-reading record at SHA-256 9ce3686a90a41af916a370b8d4ec04637f22a1d32f872180c6d8a1b790e25a0e, profile transition from 876e4cc7cc5d894f5e558ac733913e84a8b779f72c77661e89d448fd1e05ced4 to 4fe769b6f13c4a719c9d31446dd3fef413eca5d3ce1f56340aada9f99b0dce64, and manifest transition from d2b5f8f1ae7e067d257eeb24b533be1d33349446d56d361ea59f4a71472eca70 to a4fdceee4089c2812d9d77be763c1738152a63231b3f06847ea93383e4a3b327. The decision digest is 1020b28d5da8d0e823cadd352ea2236c69cbb636660a573eb5d74b8c131bc5d8. It records both the source-scoped mapping and the distinct modern numeric boundary policy without raising admission above source_scoped_public.

Stage 2A validation proves policy immutability, capability gating, solar-bound ordering, half-open boundary ownership, weekday and nominal-schedule selection, UTC-to-UT1 adapter behavior, strict transport fields, and explicit polar failure. The solar boundary is additionally authority-validated against the offline JPL Horizons sun-new-york-equinox observer-table fixture: the content-identified DE-0441LE-0441 validation run differed by 0.082 s at sunrise and 0.123 s at sunset under the fixture’s 2 s gate. That comparison names the -0.833-degree Moira versus -0.8333-degree fixture thresholds and UT1-versus-UT labels rather than claiming identical-threshold parity. It validates the astronomical boundary only; it is not an external Pancha Pakshi oracle or corroboration of the 1879 witness.

Stage 2B validation reconstructs and verifies the frozen Stage 2A manifest before checking the current manifest, then proves exact 1,440-second nazhigai arithmetic on reader-bound TT, 30-nazhigai/43,200-second closure, contiguous half-open cells, TT-to-UT1 endpoint projection, unclipped topology reporting, the 0.0001 s coalescence policy, immutable vessels, strict facade/transport policy admission, and the absence of current-cell selection or proportional solar scaling. The Stage 2A Horizons evidence continues to govern only the solar anchor; it is not an oracle for the fixed-clock composition.

Stage 2C validation reconstructs and verifies the frozen Stage 2B manifest, then proves selection at all cell midpoints and shared boundaries, inclusive anchor and excluded fixed-end ownership, exact sunset/sunrise solar-half precedence, explicit long-half tail behavior, ineligibility of prior-half cells after a short-half boundary, zero-tolerance membership, immutable status/cell consistency, strict facade/transport policy admission, and the absence of clipping, wrapping, repeating, scaling, or paksha inference. This is structural and physical-invariant evidence over admitted inputs; no external current-cell oracle or new astronomical-accuracy claim is made.

Stage 2D validation reconstructs and verifies the frozen Stage 2C manifest, then proves reduced fraction ordering and unit closure, independent common-anchor TT endpoint mapping, exact TT/UT1 outer closure, 25 positive contiguous half-open cells across long and short day and night halves, capability gating, immutable vessels, strict facade/transport policy admission, rejection of fraction/endpoint drift and contradictory routing provenance, and honest route-specific provenance. The inherited Horizons comparison still governs only the solar anchor. No external Pancha Pakshi proportional-timing oracle or new historical-accuracy claim is made.

Stage 2E validation reconstructs and verifies the frozen Stage 2D manifest, then proves exact selection at all 25 cell midpoints, the anchor, all 24 shared endpoints, immediately preceding representable instants, and the excluded old-half endpoint under solar-half-first precedence. It checks representative long and short day and night halves, zero-tolerance TT membership, exactly one non-null materialization member, immutable policy/provenance binding, capability gates, strict facade/transport admission, and fail-closed rejection of gaps, overlaps, copied cells, tail states, fixed-clock mixing, and inference. DE441 exercises the reader-bound TT/UT1 path; it is not an external Pancha Pakshi current-cell oracle or new historical-accuracy claim.

Stage 2F validation verifies both direct mapping locators, schema/hash/admission chain, exact 0 and 180-degree ownership, normalized 360 behavior, immediately adjacent representable values, one UT1-to-TT conversion shared by both body evaluations, source-label and locator consistency, immutable vessels, capability gates, and strict facade/REST input and response shapes. DE441 exercises the reader-bound astronomical path; synthetic boundary and Panchanga coherence tests establish the declared partition. No external Pancha Pakshi oracle, phase-event timing accuracy, independent witness, or universal canon is claimed.

Stage 2G validation separately verifies the Bogamuni leaf n52 Purva mapping, leaf n64 governing Amara verse and rejected commentary conflict, leaf n167 phase binding, all 54 exact mapping cells, profile/manifest hash integrity, capability separation, immutable source and composition vessels, and strict facade/REST/OpenAPI contracts. Mathematical invariants test all exact nakshatra boundaries and their adjacent representable values. A discovered, content-identified DE441 kernel exercises the apparent geocentric and reader-bound TT execution path. These are respectively source-table evidence, mathematical boundary evidence, and astronomical execution evidence; none is an external natal-identity oracle or a universal-canon proof.

Stage 2H validation separately verifies the Bogamuni n52 and n60 governing Paksha stanzas, the n157 and n158 internal repetitions, all fourteen exact Paksha-by-weekday cells, profile/manifest/decision hash integrity, immutable provenance, capability isolation, and strict facade/REST/OpenAPI contracts. Adversarial checks reject a day/night dimension, foreign locators, incomplete or duplicated cells, forged birds, RULE equivalence, first_eat_bird or Adhikara/Bharana relabelling, and temporal or scoring fields. These are source-table and structural-contract evidence, not an external condition, electional, or forecasting oracle.

The chained Stage 2I admission decision binds the unchanged 1879 profile at SHA-256 4fe769b6f13c4a719c9d31446dd3fef413eca5d3ce1f56340aada9f99b0dce64, the frozen Stage 2H decision at SHA-256 9ea7c871643bb8fc68d420223d0090ca91699154c761c67ccaf9201f401906cd, and the manifest transition from eae9fc471da08eccf24515ef12cdaf59330aa1b7ad7f9d43432c7a1482704a03 to d1aba3757910ded019cb6a2a5d6fb92c2e1ebbea755c26953dff1347834bf0e8. Its decision digest is 83c9bc0a423c09ccc113007625fee4a7d6b9ee1e890827f71595c96c3f826807. Stage 2I validation checks all 28 cells, the four governing 1879 leaves, same-witness locator retention, parity with the existing nominal schedule’s first-EAT field and initial EAT cell, immutable canonical provenance, capability isolation, and strict facade/REST/OpenAPI shapes. Uromarisi 1934 is recorded only as separate-publication corroboration with unestablished textual lineage independence; it supplies no runtime cell or universal-canon proof.

The Stage 2J research decision has canonical SHA-256 d04ed0f3716fe605dc5d8172114dc759b30c4e87be968eebc36e35a23d789243. It binds the unchanged Stage 2I manifest and decision, four inspected PDF hashes and exact page locators, the recovered five-position ordinal axis, the two exact rational selector candidates, and the fail-closed no-default and no-cross-witness-binding policy. Its regression tests also prove that no vinadi token entered the manifest, package exports, facade, or REST routes. This is source-specific structural and policy-conflict research evidence, not translation authority, external prognostic parity, or public admission.

Schema/hash/source checks are regression integrity; exact closure, bijection, partition, immutability, and no-default checks are structural invariants; the named leaf readings are source-specific evidence. None is an external oracle or proof of a universal, historically original, or most widely practised Pancha Pakshi canon.

9. Source-Artifact Boundary

Moira distributes its own code, schemas, prose, and independently normalized symbolic facts. It does not bundle archival scans, derivative PDFs, OCR, page images, copied table layouts, source prose, or third-party translations. Archive rights and license labels remain bibliographic metadata rather than runtime inputs. This is a standing non-bundling architecture, not a separate rights-clearance phase or a public-admission blocker.

10. Deferred Products

The following require separately named sources, policies, capabilities, and tests: source-attested or alternate solar-proportional doctrines and any selectors belonging to those unadmitted doctrines; alternate Padu doctrines; Bharana or Adhikara bird products; alternate Sookshma selector doctrines, any cross-witness binding of Stage 2K selectors to Uromarisi ordinal outcomes, vinadi clock routing, and semantic outcome products beyond the Stage 2J axis, Stage 2P research-only illness locator grid, Stage 2Q five-cell EAT pilot, Stage 2R five-cell WALK extension, Stage 2S five-cell RULE extension, and Stage 2T four-cell SLEEP extension with verse 250 blocked by identity conflict; condition/scoring; and electional window search. Fixed-clock materialization, fixed-clock current-cell selection, solar-proportional materialization, and solar-proportional current-cell selection are admitted only through their explicit Stage 2B, Stage 2C, Stage 2D, and Stage 2E policies and surfaces above; none may be inferred from the source-scoped profile or from the Stage 2A local-solar context alone. Astronomical paksha inference is admitted only through the separate Stage 2F policy and never supplies paksha to those operations automatically. Natal-Moon identity is admitted only through the separate Stage 2G Bogamuni profile and fixed modern composition; alternate natal doctrines remain separate future profiles rather than overrides to this one. Padu-bird lookup is admitted only through the separate Stage 2H profile and its explicit Paksha-by-weekday table; it does not supply any schedule, identity, Adhikara/Bharana alias, condition, score, or forecast. First-samam EAT lookup is admitted only through the Stage 2I capability on the 1879 profile. It does not materialize a schedule or supply an ambient eating, Padu, authority, Adhikara/Bharana, condition, score, or forecast role.

This site is open source. Improve this page.